Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
d397712b | 25 | #include <linux/buffer_head.h> |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a74a4b97 | 27 | #include <linux/kthread.h> |
4b4e25f2 | 28 | #include <linux/freezer.h> |
163e783e | 29 | #include <linux/crc32c.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
784b4e29 | 31 | #include <linux/migrate.h> |
7a36ddec | 32 | #include <linux/ratelimit.h> |
7e75bf3f | 33 | #include <asm/unaligned.h> |
4b4e25f2 | 34 | #include "compat.h" |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
8b712842 | 41 | #include "async-thread.h" |
925baedd | 42 | #include "locking.h" |
e02119d5 | 43 | #include "tree-log.h" |
fa9c0d79 | 44 | #include "free-space-cache.h" |
581bb050 | 45 | #include "inode-map.h" |
21adbd5c | 46 | #include "check-integrity.h" |
eb60ceac | 47 | |
d1310b2e | 48 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 49 | static void end_workqueue_fn(struct btrfs_work *work); |
4df27c4d | 50 | static void free_fs_root(struct btrfs_root *root); |
acce952b | 51 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
52 | int read_only); | |
53 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root); | |
54 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root); | |
55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
56 | struct btrfs_root *root); | |
57 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t); | |
58 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root); | |
59 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
60 | struct extent_io_tree *dirty_pages, | |
61 | int mark); | |
62 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
63 | struct extent_io_tree *pinned_extents); | |
64 | static int btrfs_cleanup_transaction(struct btrfs_root *root); | |
ce9adaa5 | 65 | |
d352ac68 CM |
66 | /* |
67 | * end_io_wq structs are used to do processing in task context when an IO is | |
68 | * complete. This is used during reads to verify checksums, and it is used | |
69 | * by writes to insert metadata for new file extents after IO is complete. | |
70 | */ | |
ce9adaa5 CM |
71 | struct end_io_wq { |
72 | struct bio *bio; | |
73 | bio_end_io_t *end_io; | |
74 | void *private; | |
75 | struct btrfs_fs_info *info; | |
76 | int error; | |
22c59948 | 77 | int metadata; |
ce9adaa5 | 78 | struct list_head list; |
8b712842 | 79 | struct btrfs_work work; |
ce9adaa5 | 80 | }; |
0da5468f | 81 | |
d352ac68 CM |
82 | /* |
83 | * async submit bios are used to offload expensive checksumming | |
84 | * onto the worker threads. They checksum file and metadata bios | |
85 | * just before they are sent down the IO stack. | |
86 | */ | |
44b8bd7e CM |
87 | struct async_submit_bio { |
88 | struct inode *inode; | |
89 | struct bio *bio; | |
90 | struct list_head list; | |
4a69a410 CM |
91 | extent_submit_bio_hook_t *submit_bio_start; |
92 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
93 | int rw; |
94 | int mirror_num; | |
c8b97818 | 95 | unsigned long bio_flags; |
eaf25d93 CM |
96 | /* |
97 | * bio_offset is optional, can be used if the pages in the bio | |
98 | * can't tell us where in the file the bio should go | |
99 | */ | |
100 | u64 bio_offset; | |
8b712842 | 101 | struct btrfs_work work; |
44b8bd7e CM |
102 | }; |
103 | ||
85d4e461 CM |
104 | /* |
105 | * Lockdep class keys for extent_buffer->lock's in this root. For a given | |
106 | * eb, the lockdep key is determined by the btrfs_root it belongs to and | |
107 | * the level the eb occupies in the tree. | |
108 | * | |
109 | * Different roots are used for different purposes and may nest inside each | |
110 | * other and they require separate keysets. As lockdep keys should be | |
111 | * static, assign keysets according to the purpose of the root as indicated | |
112 | * by btrfs_root->objectid. This ensures that all special purpose roots | |
113 | * have separate keysets. | |
4008c04a | 114 | * |
85d4e461 CM |
115 | * Lock-nesting across peer nodes is always done with the immediate parent |
116 | * node locked thus preventing deadlock. As lockdep doesn't know this, use | |
117 | * subclass to avoid triggering lockdep warning in such cases. | |
4008c04a | 118 | * |
85d4e461 CM |
119 | * The key is set by the readpage_end_io_hook after the buffer has passed |
120 | * csum validation but before the pages are unlocked. It is also set by | |
121 | * btrfs_init_new_buffer on freshly allocated blocks. | |
4008c04a | 122 | * |
85d4e461 CM |
123 | * We also add a check to make sure the highest level of the tree is the |
124 | * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code | |
125 | * needs update as well. | |
4008c04a CM |
126 | */ |
127 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
128 | # if BTRFS_MAX_LEVEL != 8 | |
129 | # error | |
130 | # endif | |
85d4e461 CM |
131 | |
132 | static struct btrfs_lockdep_keyset { | |
133 | u64 id; /* root objectid */ | |
134 | const char *name_stem; /* lock name stem */ | |
135 | char names[BTRFS_MAX_LEVEL + 1][20]; | |
136 | struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; | |
137 | } btrfs_lockdep_keysets[] = { | |
138 | { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, | |
139 | { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, | |
140 | { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, | |
141 | { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, | |
142 | { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, | |
143 | { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, | |
144 | { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" }, | |
145 | { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, | |
146 | { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, | |
147 | { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, | |
148 | { .id = 0, .name_stem = "tree" }, | |
4008c04a | 149 | }; |
85d4e461 CM |
150 | |
151 | void __init btrfs_init_lockdep(void) | |
152 | { | |
153 | int i, j; | |
154 | ||
155 | /* initialize lockdep class names */ | |
156 | for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { | |
157 | struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; | |
158 | ||
159 | for (j = 0; j < ARRAY_SIZE(ks->names); j++) | |
160 | snprintf(ks->names[j], sizeof(ks->names[j]), | |
161 | "btrfs-%s-%02d", ks->name_stem, j); | |
162 | } | |
163 | } | |
164 | ||
165 | void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, | |
166 | int level) | |
167 | { | |
168 | struct btrfs_lockdep_keyset *ks; | |
169 | ||
170 | BUG_ON(level >= ARRAY_SIZE(ks->keys)); | |
171 | ||
172 | /* find the matching keyset, id 0 is the default entry */ | |
173 | for (ks = btrfs_lockdep_keysets; ks->id; ks++) | |
174 | if (ks->id == objectid) | |
175 | break; | |
176 | ||
177 | lockdep_set_class_and_name(&eb->lock, | |
178 | &ks->keys[level], ks->names[level]); | |
179 | } | |
180 | ||
4008c04a CM |
181 | #endif |
182 | ||
d352ac68 CM |
183 | /* |
184 | * extents on the btree inode are pretty simple, there's one extent | |
185 | * that covers the entire device | |
186 | */ | |
b2950863 | 187 | static struct extent_map *btree_get_extent(struct inode *inode, |
306e16ce | 188 | struct page *page, size_t pg_offset, u64 start, u64 len, |
b2950863 | 189 | int create) |
7eccb903 | 190 | { |
5f39d397 CM |
191 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
192 | struct extent_map *em; | |
193 | int ret; | |
194 | ||
890871be | 195 | read_lock(&em_tree->lock); |
d1310b2e | 196 | em = lookup_extent_mapping(em_tree, start, len); |
a061fc8d CM |
197 | if (em) { |
198 | em->bdev = | |
199 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
890871be | 200 | read_unlock(&em_tree->lock); |
5f39d397 | 201 | goto out; |
a061fc8d | 202 | } |
890871be | 203 | read_unlock(&em_tree->lock); |
7b13b7b1 | 204 | |
172ddd60 | 205 | em = alloc_extent_map(); |
5f39d397 CM |
206 | if (!em) { |
207 | em = ERR_PTR(-ENOMEM); | |
208 | goto out; | |
209 | } | |
210 | em->start = 0; | |
0afbaf8c | 211 | em->len = (u64)-1; |
c8b97818 | 212 | em->block_len = (u64)-1; |
5f39d397 | 213 | em->block_start = 0; |
a061fc8d | 214 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e | 215 | |
890871be | 216 | write_lock(&em_tree->lock); |
5f39d397 CM |
217 | ret = add_extent_mapping(em_tree, em); |
218 | if (ret == -EEXIST) { | |
0afbaf8c CM |
219 | u64 failed_start = em->start; |
220 | u64 failed_len = em->len; | |
221 | ||
5f39d397 | 222 | free_extent_map(em); |
7b13b7b1 | 223 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c | 224 | if (em) { |
7b13b7b1 | 225 | ret = 0; |
0afbaf8c CM |
226 | } else { |
227 | em = lookup_extent_mapping(em_tree, failed_start, | |
228 | failed_len); | |
7b13b7b1 | 229 | ret = -EIO; |
0afbaf8c | 230 | } |
5f39d397 | 231 | } else if (ret) { |
7b13b7b1 CM |
232 | free_extent_map(em); |
233 | em = NULL; | |
5f39d397 | 234 | } |
890871be | 235 | write_unlock(&em_tree->lock); |
7b13b7b1 CM |
236 | |
237 | if (ret) | |
238 | em = ERR_PTR(ret); | |
5f39d397 CM |
239 | out: |
240 | return em; | |
7eccb903 CM |
241 | } |
242 | ||
19c00ddc CM |
243 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
244 | { | |
163e783e | 245 | return crc32c(seed, data, len); |
19c00ddc CM |
246 | } |
247 | ||
248 | void btrfs_csum_final(u32 crc, char *result) | |
249 | { | |
7e75bf3f | 250 | put_unaligned_le32(~crc, result); |
19c00ddc CM |
251 | } |
252 | ||
d352ac68 CM |
253 | /* |
254 | * compute the csum for a btree block, and either verify it or write it | |
255 | * into the csum field of the block. | |
256 | */ | |
19c00ddc CM |
257 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
258 | int verify) | |
259 | { | |
6c41761f | 260 | u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); |
607d432d | 261 | char *result = NULL; |
19c00ddc CM |
262 | unsigned long len; |
263 | unsigned long cur_len; | |
264 | unsigned long offset = BTRFS_CSUM_SIZE; | |
19c00ddc CM |
265 | char *kaddr; |
266 | unsigned long map_start; | |
267 | unsigned long map_len; | |
268 | int err; | |
269 | u32 crc = ~(u32)0; | |
607d432d | 270 | unsigned long inline_result; |
19c00ddc CM |
271 | |
272 | len = buf->len - offset; | |
d397712b | 273 | while (len > 0) { |
19c00ddc | 274 | err = map_private_extent_buffer(buf, offset, 32, |
a6591715 | 275 | &kaddr, &map_start, &map_len); |
d397712b | 276 | if (err) |
19c00ddc | 277 | return 1; |
19c00ddc CM |
278 | cur_len = min(len, map_len - (offset - map_start)); |
279 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
280 | crc, cur_len); | |
281 | len -= cur_len; | |
282 | offset += cur_len; | |
19c00ddc | 283 | } |
607d432d JB |
284 | if (csum_size > sizeof(inline_result)) { |
285 | result = kzalloc(csum_size * sizeof(char), GFP_NOFS); | |
286 | if (!result) | |
287 | return 1; | |
288 | } else { | |
289 | result = (char *)&inline_result; | |
290 | } | |
291 | ||
19c00ddc CM |
292 | btrfs_csum_final(crc, result); |
293 | ||
294 | if (verify) { | |
607d432d | 295 | if (memcmp_extent_buffer(buf, result, 0, csum_size)) { |
e4204ded CM |
296 | u32 val; |
297 | u32 found = 0; | |
607d432d | 298 | memcpy(&found, result, csum_size); |
e4204ded | 299 | |
607d432d | 300 | read_extent_buffer(buf, &val, 0, csum_size); |
7a36ddec | 301 | printk_ratelimited(KERN_INFO "btrfs: %s checksum verify " |
193f284d CM |
302 | "failed on %llu wanted %X found %X " |
303 | "level %d\n", | |
304 | root->fs_info->sb->s_id, | |
305 | (unsigned long long)buf->start, val, found, | |
306 | btrfs_header_level(buf)); | |
607d432d JB |
307 | if (result != (char *)&inline_result) |
308 | kfree(result); | |
19c00ddc CM |
309 | return 1; |
310 | } | |
311 | } else { | |
607d432d | 312 | write_extent_buffer(buf, result, 0, csum_size); |
19c00ddc | 313 | } |
607d432d JB |
314 | if (result != (char *)&inline_result) |
315 | kfree(result); | |
19c00ddc CM |
316 | return 0; |
317 | } | |
318 | ||
d352ac68 CM |
319 | /* |
320 | * we can't consider a given block up to date unless the transid of the | |
321 | * block matches the transid in the parent node's pointer. This is how we | |
322 | * detect blocks that either didn't get written at all or got written | |
323 | * in the wrong place. | |
324 | */ | |
1259ab75 CM |
325 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
326 | struct extent_buffer *eb, u64 parent_transid) | |
327 | { | |
2ac55d41 | 328 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
329 | int ret; |
330 | ||
331 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
332 | return 0; | |
333 | ||
2ac55d41 JB |
334 | lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, |
335 | 0, &cached_state, GFP_NOFS); | |
336 | if (extent_buffer_uptodate(io_tree, eb, cached_state) && | |
1259ab75 CM |
337 | btrfs_header_generation(eb) == parent_transid) { |
338 | ret = 0; | |
339 | goto out; | |
340 | } | |
7a36ddec | 341 | printk_ratelimited("parent transid verify failed on %llu wanted %llu " |
193f284d CM |
342 | "found %llu\n", |
343 | (unsigned long long)eb->start, | |
344 | (unsigned long long)parent_transid, | |
345 | (unsigned long long)btrfs_header_generation(eb)); | |
1259ab75 | 346 | ret = 1; |
2ac55d41 | 347 | clear_extent_buffer_uptodate(io_tree, eb, &cached_state); |
33958dc6 | 348 | out: |
2ac55d41 JB |
349 | unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, |
350 | &cached_state, GFP_NOFS); | |
1259ab75 | 351 | return ret; |
1259ab75 CM |
352 | } |
353 | ||
d352ac68 CM |
354 | /* |
355 | * helper to read a given tree block, doing retries as required when | |
356 | * the checksums don't match and we have alternate mirrors to try. | |
357 | */ | |
f188591e CM |
358 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
359 | struct extent_buffer *eb, | |
ca7a79ad | 360 | u64 start, u64 parent_transid) |
f188591e CM |
361 | { |
362 | struct extent_io_tree *io_tree; | |
363 | int ret; | |
364 | int num_copies = 0; | |
365 | int mirror_num = 0; | |
366 | ||
a826d6dc | 367 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
f188591e CM |
368 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
369 | while (1) { | |
bb82ab88 AJ |
370 | ret = read_extent_buffer_pages(io_tree, eb, start, |
371 | WAIT_COMPLETE, | |
f188591e | 372 | btree_get_extent, mirror_num); |
1259ab75 CM |
373 | if (!ret && |
374 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 375 | return ret; |
d397712b | 376 | |
a826d6dc JB |
377 | /* |
378 | * This buffer's crc is fine, but its contents are corrupted, so | |
379 | * there is no reason to read the other copies, they won't be | |
380 | * any less wrong. | |
381 | */ | |
382 | if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) | |
383 | return ret; | |
384 | ||
f188591e CM |
385 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
386 | eb->start, eb->len); | |
4235298e | 387 | if (num_copies == 1) |
f188591e | 388 | return ret; |
4235298e | 389 | |
f188591e | 390 | mirror_num++; |
4235298e | 391 | if (mirror_num > num_copies) |
f188591e | 392 | return ret; |
f188591e | 393 | } |
f188591e CM |
394 | return -EIO; |
395 | } | |
19c00ddc | 396 | |
d352ac68 | 397 | /* |
d397712b CM |
398 | * checksum a dirty tree block before IO. This has extra checks to make sure |
399 | * we only fill in the checksum field in the first page of a multi-page block | |
d352ac68 | 400 | */ |
d397712b | 401 | |
b2950863 | 402 | static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
19c00ddc | 403 | { |
d1310b2e | 404 | struct extent_io_tree *tree; |
35ebb934 | 405 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc | 406 | u64 found_start; |
19c00ddc CM |
407 | unsigned long len; |
408 | struct extent_buffer *eb; | |
f188591e CM |
409 | int ret; |
410 | ||
d1310b2e | 411 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc | 412 | |
eb14ab8e CM |
413 | if (page->private == EXTENT_PAGE_PRIVATE) { |
414 | WARN_ON(1); | |
19c00ddc | 415 | goto out; |
eb14ab8e CM |
416 | } |
417 | if (!page->private) { | |
418 | WARN_ON(1); | |
19c00ddc | 419 | goto out; |
eb14ab8e | 420 | } |
19c00ddc | 421 | len = page->private >> 2; |
d397712b CM |
422 | WARN_ON(len == 0); |
423 | ||
ba144192 | 424 | eb = alloc_extent_buffer(tree, start, len, page); |
91ca338d TI |
425 | if (eb == NULL) { |
426 | WARN_ON(1); | |
427 | goto out; | |
428 | } | |
ca7a79ad CM |
429 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
430 | btrfs_header_generation(eb)); | |
f188591e | 431 | BUG_ON(ret); |
784b4e29 CM |
432 | WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN)); |
433 | ||
19c00ddc CM |
434 | found_start = btrfs_header_bytenr(eb); |
435 | if (found_start != start) { | |
55c69072 CM |
436 | WARN_ON(1); |
437 | goto err; | |
438 | } | |
439 | if (eb->first_page != page) { | |
55c69072 CM |
440 | WARN_ON(1); |
441 | goto err; | |
442 | } | |
443 | if (!PageUptodate(page)) { | |
55c69072 CM |
444 | WARN_ON(1); |
445 | goto err; | |
19c00ddc | 446 | } |
19c00ddc | 447 | csum_tree_block(root, eb, 0); |
55c69072 | 448 | err: |
19c00ddc CM |
449 | free_extent_buffer(eb); |
450 | out: | |
451 | return 0; | |
452 | } | |
453 | ||
2b82032c YZ |
454 | static int check_tree_block_fsid(struct btrfs_root *root, |
455 | struct extent_buffer *eb) | |
456 | { | |
457 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
458 | u8 fsid[BTRFS_UUID_SIZE]; | |
459 | int ret = 1; | |
460 | ||
461 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
462 | BTRFS_FSID_SIZE); | |
463 | while (fs_devices) { | |
464 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
465 | ret = 0; | |
466 | break; | |
467 | } | |
468 | fs_devices = fs_devices->seed; | |
469 | } | |
470 | return ret; | |
471 | } | |
472 | ||
a826d6dc JB |
473 | #define CORRUPT(reason, eb, root, slot) \ |
474 | printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \ | |
475 | "root=%llu, slot=%d\n", reason, \ | |
476 | (unsigned long long)btrfs_header_bytenr(eb), \ | |
477 | (unsigned long long)root->objectid, slot) | |
478 | ||
479 | static noinline int check_leaf(struct btrfs_root *root, | |
480 | struct extent_buffer *leaf) | |
481 | { | |
482 | struct btrfs_key key; | |
483 | struct btrfs_key leaf_key; | |
484 | u32 nritems = btrfs_header_nritems(leaf); | |
485 | int slot; | |
486 | ||
487 | if (nritems == 0) | |
488 | return 0; | |
489 | ||
490 | /* Check the 0 item */ | |
491 | if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != | |
492 | BTRFS_LEAF_DATA_SIZE(root)) { | |
493 | CORRUPT("invalid item offset size pair", leaf, root, 0); | |
494 | return -EIO; | |
495 | } | |
496 | ||
497 | /* | |
498 | * Check to make sure each items keys are in the correct order and their | |
499 | * offsets make sense. We only have to loop through nritems-1 because | |
500 | * we check the current slot against the next slot, which verifies the | |
501 | * next slot's offset+size makes sense and that the current's slot | |
502 | * offset is correct. | |
503 | */ | |
504 | for (slot = 0; slot < nritems - 1; slot++) { | |
505 | btrfs_item_key_to_cpu(leaf, &leaf_key, slot); | |
506 | btrfs_item_key_to_cpu(leaf, &key, slot + 1); | |
507 | ||
508 | /* Make sure the keys are in the right order */ | |
509 | if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { | |
510 | CORRUPT("bad key order", leaf, root, slot); | |
511 | return -EIO; | |
512 | } | |
513 | ||
514 | /* | |
515 | * Make sure the offset and ends are right, remember that the | |
516 | * item data starts at the end of the leaf and grows towards the | |
517 | * front. | |
518 | */ | |
519 | if (btrfs_item_offset_nr(leaf, slot) != | |
520 | btrfs_item_end_nr(leaf, slot + 1)) { | |
521 | CORRUPT("slot offset bad", leaf, root, slot); | |
522 | return -EIO; | |
523 | } | |
524 | ||
525 | /* | |
526 | * Check to make sure that we don't point outside of the leaf, | |
527 | * just incase all the items are consistent to eachother, but | |
528 | * all point outside of the leaf. | |
529 | */ | |
530 | if (btrfs_item_end_nr(leaf, slot) > | |
531 | BTRFS_LEAF_DATA_SIZE(root)) { | |
532 | CORRUPT("slot end outside of leaf", leaf, root, slot); | |
533 | return -EIO; | |
534 | } | |
535 | } | |
536 | ||
537 | return 0; | |
538 | } | |
539 | ||
b2950863 | 540 | static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
ce9adaa5 CM |
541 | struct extent_state *state) |
542 | { | |
543 | struct extent_io_tree *tree; | |
544 | u64 found_start; | |
545 | int found_level; | |
546 | unsigned long len; | |
547 | struct extent_buffer *eb; | |
548 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 549 | int ret = 0; |
ce9adaa5 CM |
550 | |
551 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
552 | if (page->private == EXTENT_PAGE_PRIVATE) | |
553 | goto out; | |
554 | if (!page->private) | |
555 | goto out; | |
d397712b | 556 | |
ce9adaa5 | 557 | len = page->private >> 2; |
d397712b CM |
558 | WARN_ON(len == 0); |
559 | ||
ba144192 | 560 | eb = alloc_extent_buffer(tree, start, len, page); |
91ca338d TI |
561 | if (eb == NULL) { |
562 | ret = -EIO; | |
563 | goto out; | |
564 | } | |
f188591e | 565 | |
ce9adaa5 | 566 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 567 | if (found_start != start) { |
7a36ddec | 568 | printk_ratelimited(KERN_INFO "btrfs bad tree block start " |
193f284d CM |
569 | "%llu %llu\n", |
570 | (unsigned long long)found_start, | |
571 | (unsigned long long)eb->start); | |
f188591e | 572 | ret = -EIO; |
ce9adaa5 CM |
573 | goto err; |
574 | } | |
575 | if (eb->first_page != page) { | |
d397712b CM |
576 | printk(KERN_INFO "btrfs bad first page %lu %lu\n", |
577 | eb->first_page->index, page->index); | |
ce9adaa5 | 578 | WARN_ON(1); |
f188591e | 579 | ret = -EIO; |
ce9adaa5 CM |
580 | goto err; |
581 | } | |
2b82032c | 582 | if (check_tree_block_fsid(root, eb)) { |
7a36ddec | 583 | printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n", |
193f284d | 584 | (unsigned long long)eb->start); |
1259ab75 CM |
585 | ret = -EIO; |
586 | goto err; | |
587 | } | |
ce9adaa5 CM |
588 | found_level = btrfs_header_level(eb); |
589 | ||
85d4e461 CM |
590 | btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), |
591 | eb, found_level); | |
4008c04a | 592 | |
ce9adaa5 | 593 | ret = csum_tree_block(root, eb, 1); |
a826d6dc | 594 | if (ret) { |
f188591e | 595 | ret = -EIO; |
a826d6dc JB |
596 | goto err; |
597 | } | |
598 | ||
599 | /* | |
600 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
601 | * that we don't try and read the other copies of this block, just | |
602 | * return -EIO. | |
603 | */ | |
604 | if (found_level == 0 && check_leaf(root, eb)) { | |
605 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); | |
606 | ret = -EIO; | |
607 | } | |
ce9adaa5 CM |
608 | |
609 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
610 | end = eb->start + end - 1; | |
ce9adaa5 | 611 | err: |
4bb31e92 AJ |
612 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { |
613 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
614 | btree_readahead_hook(root, eb, eb->start, ret); | |
615 | } | |
616 | ||
ce9adaa5 CM |
617 | free_extent_buffer(eb); |
618 | out: | |
f188591e | 619 | return ret; |
ce9adaa5 CM |
620 | } |
621 | ||
4bb31e92 AJ |
622 | static int btree_io_failed_hook(struct bio *failed_bio, |
623 | struct page *page, u64 start, u64 end, | |
32240a91 | 624 | int mirror_num, struct extent_state *state) |
4bb31e92 AJ |
625 | { |
626 | struct extent_io_tree *tree; | |
627 | unsigned long len; | |
628 | struct extent_buffer *eb; | |
629 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
630 | ||
631 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
632 | if (page->private == EXTENT_PAGE_PRIVATE) | |
633 | goto out; | |
634 | if (!page->private) | |
635 | goto out; | |
636 | ||
637 | len = page->private >> 2; | |
638 | WARN_ON(len == 0); | |
639 | ||
640 | eb = alloc_extent_buffer(tree, start, len, page); | |
641 | if (eb == NULL) | |
642 | goto out; | |
643 | ||
644 | if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) { | |
645 | clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags); | |
646 | btree_readahead_hook(root, eb, eb->start, -EIO); | |
647 | } | |
c674e04e | 648 | free_extent_buffer(eb); |
4bb31e92 AJ |
649 | |
650 | out: | |
651 | return -EIO; /* we fixed nothing */ | |
652 | } | |
653 | ||
ce9adaa5 | 654 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
655 | { |
656 | struct end_io_wq *end_io_wq = bio->bi_private; | |
657 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 658 | |
ce9adaa5 | 659 | fs_info = end_io_wq->info; |
ce9adaa5 | 660 | end_io_wq->error = err; |
8b712842 CM |
661 | end_io_wq->work.func = end_workqueue_fn; |
662 | end_io_wq->work.flags = 0; | |
d20f7043 | 663 | |
7b6d91da | 664 | if (bio->bi_rw & REQ_WRITE) { |
0cb59c99 | 665 | if (end_io_wq->metadata == 1) |
cad321ad CM |
666 | btrfs_queue_worker(&fs_info->endio_meta_write_workers, |
667 | &end_io_wq->work); | |
0cb59c99 JB |
668 | else if (end_io_wq->metadata == 2) |
669 | btrfs_queue_worker(&fs_info->endio_freespace_worker, | |
670 | &end_io_wq->work); | |
cad321ad CM |
671 | else |
672 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
673 | &end_io_wq->work); | |
d20f7043 CM |
674 | } else { |
675 | if (end_io_wq->metadata) | |
676 | btrfs_queue_worker(&fs_info->endio_meta_workers, | |
677 | &end_io_wq->work); | |
678 | else | |
679 | btrfs_queue_worker(&fs_info->endio_workers, | |
680 | &end_io_wq->work); | |
681 | } | |
ce9adaa5 CM |
682 | } |
683 | ||
0cb59c99 JB |
684 | /* |
685 | * For the metadata arg you want | |
686 | * | |
687 | * 0 - if data | |
688 | * 1 - if normal metadta | |
689 | * 2 - if writing to the free space cache area | |
690 | */ | |
22c59948 CM |
691 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
692 | int metadata) | |
0b86a832 | 693 | { |
ce9adaa5 | 694 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
695 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
696 | if (!end_io_wq) | |
697 | return -ENOMEM; | |
698 | ||
699 | end_io_wq->private = bio->bi_private; | |
700 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 701 | end_io_wq->info = info; |
ce9adaa5 CM |
702 | end_io_wq->error = 0; |
703 | end_io_wq->bio = bio; | |
22c59948 | 704 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
705 | |
706 | bio->bi_private = end_io_wq; | |
707 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
708 | return 0; |
709 | } | |
710 | ||
b64a2851 | 711 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 712 | { |
4854ddd0 CM |
713 | unsigned long limit = min_t(unsigned long, |
714 | info->workers.max_workers, | |
715 | info->fs_devices->open_devices); | |
716 | return 256 * limit; | |
717 | } | |
0986fe9e | 718 | |
4a69a410 CM |
719 | static void run_one_async_start(struct btrfs_work *work) |
720 | { | |
4a69a410 CM |
721 | struct async_submit_bio *async; |
722 | ||
723 | async = container_of(work, struct async_submit_bio, work); | |
4a69a410 | 724 | async->submit_bio_start(async->inode, async->rw, async->bio, |
eaf25d93 CM |
725 | async->mirror_num, async->bio_flags, |
726 | async->bio_offset); | |
4a69a410 CM |
727 | } |
728 | ||
729 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
730 | { |
731 | struct btrfs_fs_info *fs_info; | |
732 | struct async_submit_bio *async; | |
4854ddd0 | 733 | int limit; |
8b712842 CM |
734 | |
735 | async = container_of(work, struct async_submit_bio, work); | |
736 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 737 | |
b64a2851 | 738 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
739 | limit = limit * 2 / 3; |
740 | ||
8b712842 | 741 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 742 | |
b64a2851 CM |
743 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
744 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
745 | wake_up(&fs_info->async_submit_wait); |
746 | ||
4a69a410 | 747 | async->submit_bio_done(async->inode, async->rw, async->bio, |
eaf25d93 CM |
748 | async->mirror_num, async->bio_flags, |
749 | async->bio_offset); | |
4a69a410 CM |
750 | } |
751 | ||
752 | static void run_one_async_free(struct btrfs_work *work) | |
753 | { | |
754 | struct async_submit_bio *async; | |
755 | ||
756 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
757 | kfree(async); |
758 | } | |
759 | ||
44b8bd7e CM |
760 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
761 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 762 | unsigned long bio_flags, |
eaf25d93 | 763 | u64 bio_offset, |
4a69a410 CM |
764 | extent_submit_bio_hook_t *submit_bio_start, |
765 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
766 | { |
767 | struct async_submit_bio *async; | |
768 | ||
769 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
770 | if (!async) | |
771 | return -ENOMEM; | |
772 | ||
773 | async->inode = inode; | |
774 | async->rw = rw; | |
775 | async->bio = bio; | |
776 | async->mirror_num = mirror_num; | |
4a69a410 CM |
777 | async->submit_bio_start = submit_bio_start; |
778 | async->submit_bio_done = submit_bio_done; | |
779 | ||
780 | async->work.func = run_one_async_start; | |
781 | async->work.ordered_func = run_one_async_done; | |
782 | async->work.ordered_free = run_one_async_free; | |
783 | ||
8b712842 | 784 | async->work.flags = 0; |
c8b97818 | 785 | async->bio_flags = bio_flags; |
eaf25d93 | 786 | async->bio_offset = bio_offset; |
8c8bee1d | 787 | |
cb03c743 | 788 | atomic_inc(&fs_info->nr_async_submits); |
d313d7a3 | 789 | |
7b6d91da | 790 | if (rw & REQ_SYNC) |
d313d7a3 CM |
791 | btrfs_set_work_high_prio(&async->work); |
792 | ||
8b712842 | 793 | btrfs_queue_worker(&fs_info->workers, &async->work); |
9473f16c | 794 | |
d397712b | 795 | while (atomic_read(&fs_info->async_submit_draining) && |
771ed689 CM |
796 | atomic_read(&fs_info->nr_async_submits)) { |
797 | wait_event(fs_info->async_submit_wait, | |
798 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
799 | } | |
800 | ||
44b8bd7e CM |
801 | return 0; |
802 | } | |
803 | ||
ce3ed71a CM |
804 | static int btree_csum_one_bio(struct bio *bio) |
805 | { | |
806 | struct bio_vec *bvec = bio->bi_io_vec; | |
807 | int bio_index = 0; | |
808 | struct btrfs_root *root; | |
809 | ||
810 | WARN_ON(bio->bi_vcnt <= 0); | |
d397712b | 811 | while (bio_index < bio->bi_vcnt) { |
ce3ed71a CM |
812 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
813 | csum_dirty_buffer(root, bvec->bv_page); | |
814 | bio_index++; | |
815 | bvec++; | |
816 | } | |
817 | return 0; | |
818 | } | |
819 | ||
4a69a410 CM |
820 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
821 | struct bio *bio, int mirror_num, | |
eaf25d93 CM |
822 | unsigned long bio_flags, |
823 | u64 bio_offset) | |
22c59948 | 824 | { |
8b712842 CM |
825 | /* |
826 | * when we're called for a write, we're already in the async | |
5443be45 | 827 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 828 | */ |
4a69a410 CM |
829 | btree_csum_one_bio(bio); |
830 | return 0; | |
831 | } | |
22c59948 | 832 | |
4a69a410 | 833 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
834 | int mirror_num, unsigned long bio_flags, |
835 | u64 bio_offset) | |
4a69a410 | 836 | { |
8b712842 | 837 | /* |
4a69a410 CM |
838 | * when we're called for a write, we're already in the async |
839 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 840 | */ |
8b712842 | 841 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
842 | } |
843 | ||
44b8bd7e | 844 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
eaf25d93 CM |
845 | int mirror_num, unsigned long bio_flags, |
846 | u64 bio_offset) | |
44b8bd7e | 847 | { |
cad321ad CM |
848 | int ret; |
849 | ||
850 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
851 | bio, 1); | |
852 | BUG_ON(ret); | |
853 | ||
7b6d91da | 854 | if (!(rw & REQ_WRITE)) { |
4a69a410 CM |
855 | /* |
856 | * called for a read, do the setup so that checksum validation | |
857 | * can happen in the async kernel threads | |
858 | */ | |
4a69a410 | 859 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
6f3577bd | 860 | mirror_num, 0); |
44b8bd7e | 861 | } |
d313d7a3 | 862 | |
cad321ad CM |
863 | /* |
864 | * kthread helpers are used to submit writes so that checksumming | |
865 | * can happen in parallel across all CPUs | |
866 | */ | |
44b8bd7e | 867 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
c8b97818 | 868 | inode, rw, bio, mirror_num, 0, |
eaf25d93 | 869 | bio_offset, |
4a69a410 CM |
870 | __btree_submit_bio_start, |
871 | __btree_submit_bio_done); | |
44b8bd7e CM |
872 | } |
873 | ||
3dd1462e | 874 | #ifdef CONFIG_MIGRATION |
784b4e29 CM |
875 | static int btree_migratepage(struct address_space *mapping, |
876 | struct page *newpage, struct page *page) | |
877 | { | |
878 | /* | |
879 | * we can't safely write a btree page from here, | |
880 | * we haven't done the locking hook | |
881 | */ | |
882 | if (PageDirty(page)) | |
883 | return -EAGAIN; | |
884 | /* | |
885 | * Buffers may be managed in a filesystem specific way. | |
886 | * We must have no buffers or drop them. | |
887 | */ | |
888 | if (page_has_private(page) && | |
889 | !try_to_release_page(page, GFP_KERNEL)) | |
890 | return -EAGAIN; | |
784b4e29 CM |
891 | return migrate_page(mapping, newpage, page); |
892 | } | |
3dd1462e | 893 | #endif |
784b4e29 | 894 | |
0da5468f CM |
895 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
896 | { | |
d1310b2e | 897 | struct extent_io_tree *tree; |
b9473439 CM |
898 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
899 | struct extent_buffer *eb; | |
900 | int was_dirty; | |
901 | ||
d1310b2e | 902 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
b9473439 CM |
903 | if (!(current->flags & PF_MEMALLOC)) { |
904 | return extent_write_full_page(tree, page, | |
905 | btree_get_extent, wbc); | |
906 | } | |
5443be45 | 907 | |
b9473439 | 908 | redirty_page_for_writepage(wbc, page); |
784b4e29 | 909 | eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE); |
b9473439 CM |
910 | WARN_ON(!eb); |
911 | ||
912 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); | |
913 | if (!was_dirty) { | |
914 | spin_lock(&root->fs_info->delalloc_lock); | |
915 | root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE; | |
916 | spin_unlock(&root->fs_info->delalloc_lock); | |
5443be45 | 917 | } |
b9473439 CM |
918 | free_extent_buffer(eb); |
919 | ||
920 | unlock_page(page); | |
921 | return 0; | |
5f39d397 | 922 | } |
0da5468f CM |
923 | |
924 | static int btree_writepages(struct address_space *mapping, | |
925 | struct writeback_control *wbc) | |
926 | { | |
d1310b2e CM |
927 | struct extent_io_tree *tree; |
928 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 929 | if (wbc->sync_mode == WB_SYNC_NONE) { |
b9473439 | 930 | struct btrfs_root *root = BTRFS_I(mapping->host)->root; |
793955bc | 931 | u64 num_dirty; |
24ab9cd8 | 932 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
933 | |
934 | if (wbc->for_kupdate) | |
935 | return 0; | |
936 | ||
b9473439 CM |
937 | /* this is a bit racy, but that's ok */ |
938 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
d397712b | 939 | if (num_dirty < thresh) |
793955bc | 940 | return 0; |
793955bc | 941 | } |
0da5468f CM |
942 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
943 | } | |
944 | ||
b2950863 | 945 | static int btree_readpage(struct file *file, struct page *page) |
5f39d397 | 946 | { |
d1310b2e CM |
947 | struct extent_io_tree *tree; |
948 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
8ddc7d9c | 949 | return extent_read_full_page(tree, page, btree_get_extent, 0); |
5f39d397 | 950 | } |
22b0ebda | 951 | |
70dec807 | 952 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 953 | { |
d1310b2e CM |
954 | struct extent_io_tree *tree; |
955 | struct extent_map_tree *map; | |
5f39d397 | 956 | int ret; |
d98237b3 | 957 | |
98509cfc | 958 | if (PageWriteback(page) || PageDirty(page)) |
d397712b | 959 | return 0; |
98509cfc | 960 | |
d1310b2e CM |
961 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
962 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 963 | |
0c4e538b DS |
964 | /* |
965 | * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing | |
966 | * slab allocation from alloc_extent_state down the callchain where | |
967 | * it'd hit a BUG_ON as those flags are not allowed. | |
968 | */ | |
969 | gfp_flags &= ~GFP_SLAB_BUG_MASK; | |
970 | ||
7b13b7b1 | 971 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
d397712b | 972 | if (!ret) |
6af118ce | 973 | return 0; |
6af118ce CM |
974 | |
975 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
976 | if (ret == 1) { |
977 | ClearPagePrivate(page); | |
978 | set_page_private(page, 0); | |
979 | page_cache_release(page); | |
980 | } | |
6af118ce | 981 | |
d98237b3 CM |
982 | return ret; |
983 | } | |
984 | ||
5f39d397 | 985 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 986 | { |
d1310b2e CM |
987 | struct extent_io_tree *tree; |
988 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
989 | extent_invalidatepage(tree, page, offset); |
990 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 991 | if (PagePrivate(page)) { |
d397712b CM |
992 | printk(KERN_WARNING "btrfs warning page private not zero " |
993 | "on page %llu\n", (unsigned long long)page_offset(page)); | |
9ad6b7bc CM |
994 | ClearPagePrivate(page); |
995 | set_page_private(page, 0); | |
996 | page_cache_release(page); | |
997 | } | |
d98237b3 CM |
998 | } |
999 | ||
7f09410b | 1000 | static const struct address_space_operations btree_aops = { |
d98237b3 CM |
1001 | .readpage = btree_readpage, |
1002 | .writepage = btree_writepage, | |
0da5468f | 1003 | .writepages = btree_writepages, |
5f39d397 CM |
1004 | .releasepage = btree_releasepage, |
1005 | .invalidatepage = btree_invalidatepage, | |
5a92bc88 | 1006 | #ifdef CONFIG_MIGRATION |
784b4e29 | 1007 | .migratepage = btree_migratepage, |
5a92bc88 | 1008 | #endif |
d98237b3 CM |
1009 | }; |
1010 | ||
ca7a79ad CM |
1011 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1012 | u64 parent_transid) | |
090d1875 | 1013 | { |
5f39d397 CM |
1014 | struct extent_buffer *buf = NULL; |
1015 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 1016 | int ret = 0; |
090d1875 | 1017 | |
db94535d | 1018 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 1019 | if (!buf) |
090d1875 | 1020 | return 0; |
d1310b2e | 1021 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
bb82ab88 | 1022 | buf, 0, WAIT_NONE, btree_get_extent, 0); |
5f39d397 | 1023 | free_extent_buffer(buf); |
de428b63 | 1024 | return ret; |
090d1875 CM |
1025 | } |
1026 | ||
ab0fff03 AJ |
1027 | int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
1028 | int mirror_num, struct extent_buffer **eb) | |
1029 | { | |
1030 | struct extent_buffer *buf = NULL; | |
1031 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1032 | struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; | |
1033 | int ret; | |
1034 | ||
1035 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
1036 | if (!buf) | |
1037 | return 0; | |
1038 | ||
1039 | set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); | |
1040 | ||
1041 | ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, | |
1042 | btree_get_extent, mirror_num); | |
1043 | if (ret) { | |
1044 | free_extent_buffer(buf); | |
1045 | return ret; | |
1046 | } | |
1047 | ||
1048 | if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { | |
1049 | free_extent_buffer(buf); | |
1050 | return -EIO; | |
1051 | } else if (extent_buffer_uptodate(io_tree, buf, NULL)) { | |
1052 | *eb = buf; | |
1053 | } else { | |
1054 | free_extent_buffer(buf); | |
1055 | } | |
1056 | return 0; | |
1057 | } | |
1058 | ||
0999df54 CM |
1059 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
1060 | u64 bytenr, u32 blocksize) | |
1061 | { | |
1062 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1063 | struct extent_buffer *eb; | |
1064 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
f09d1f60 | 1065 | bytenr, blocksize); |
0999df54 CM |
1066 | return eb; |
1067 | } | |
1068 | ||
1069 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
1070 | u64 bytenr, u32 blocksize) | |
1071 | { | |
1072 | struct inode *btree_inode = root->fs_info->btree_inode; | |
1073 | struct extent_buffer *eb; | |
1074 | ||
1075 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
ba144192 | 1076 | bytenr, blocksize, NULL); |
0999df54 CM |
1077 | return eb; |
1078 | } | |
1079 | ||
1080 | ||
e02119d5 CM |
1081 | int btrfs_write_tree_block(struct extent_buffer *buf) |
1082 | { | |
8aa38c31 CH |
1083 | return filemap_fdatawrite_range(buf->first_page->mapping, buf->start, |
1084 | buf->start + buf->len - 1); | |
e02119d5 CM |
1085 | } |
1086 | ||
1087 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
1088 | { | |
8aa38c31 CH |
1089 | return filemap_fdatawait_range(buf->first_page->mapping, |
1090 | buf->start, buf->start + buf->len - 1); | |
e02119d5 CM |
1091 | } |
1092 | ||
0999df54 | 1093 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 1094 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
1095 | { |
1096 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
1097 | int ret; |
1098 | ||
0999df54 CM |
1099 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
1100 | if (!buf) | |
1101 | return NULL; | |
0999df54 | 1102 | |
ca7a79ad | 1103 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 | 1104 | |
d397712b | 1105 | if (ret == 0) |
b4ce94de | 1106 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
5f39d397 | 1107 | return buf; |
ce9adaa5 | 1108 | |
eb60ceac CM |
1109 | } |
1110 | ||
e089f05c | 1111 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 1112 | struct extent_buffer *buf) |
ed2ff2cb | 1113 | { |
5f39d397 | 1114 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 1115 | if (btrfs_header_generation(buf) == |
925baedd | 1116 | root->fs_info->running_transaction->transid) { |
b9447ef8 | 1117 | btrfs_assert_tree_locked(buf); |
b4ce94de | 1118 | |
b9473439 CM |
1119 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
1120 | spin_lock(&root->fs_info->delalloc_lock); | |
1121 | if (root->fs_info->dirty_metadata_bytes >= buf->len) | |
1122 | root->fs_info->dirty_metadata_bytes -= buf->len; | |
1123 | else | |
1124 | WARN_ON(1); | |
1125 | spin_unlock(&root->fs_info->delalloc_lock); | |
1126 | } | |
b4ce94de | 1127 | |
b9473439 CM |
1128 | /* ugh, clear_extent_buffer_dirty needs to lock the page */ |
1129 | btrfs_set_lock_blocking(buf); | |
d1310b2e | 1130 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 1131 | buf); |
925baedd | 1132 | } |
5f39d397 CM |
1133 | return 0; |
1134 | } | |
1135 | ||
db94535d | 1136 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 1137 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 1138 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 1139 | u64 objectid) |
d97e63b6 | 1140 | { |
cfaa7295 | 1141 | root->node = NULL; |
a28ec197 | 1142 | root->commit_root = NULL; |
db94535d CM |
1143 | root->sectorsize = sectorsize; |
1144 | root->nodesize = nodesize; | |
1145 | root->leafsize = leafsize; | |
87ee04eb | 1146 | root->stripesize = stripesize; |
123abc88 | 1147 | root->ref_cows = 0; |
0b86a832 | 1148 | root->track_dirty = 0; |
c71bf099 | 1149 | root->in_radix = 0; |
d68fc57b YZ |
1150 | root->orphan_item_inserted = 0; |
1151 | root->orphan_cleanup_state = 0; | |
0b86a832 | 1152 | |
9f5fae2f | 1153 | root->fs_info = fs_info; |
0f7d52f4 CM |
1154 | root->objectid = objectid; |
1155 | root->last_trans = 0; | |
13a8a7c8 | 1156 | root->highest_objectid = 0; |
58176a96 | 1157 | root->name = NULL; |
6bef4d31 | 1158 | root->inode_tree = RB_ROOT; |
16cdcec7 | 1159 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
f0486c68 | 1160 | root->block_rsv = NULL; |
d68fc57b | 1161 | root->orphan_block_rsv = NULL; |
0b86a832 CM |
1162 | |
1163 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 1164 | INIT_LIST_HEAD(&root->orphan_list); |
5d4f98a2 | 1165 | INIT_LIST_HEAD(&root->root_list); |
d68fc57b | 1166 | spin_lock_init(&root->orphan_lock); |
5d4f98a2 | 1167 | spin_lock_init(&root->inode_lock); |
f0486c68 | 1168 | spin_lock_init(&root->accounting_lock); |
a2135011 | 1169 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1170 | mutex_init(&root->log_mutex); |
7237f183 YZ |
1171 | init_waitqueue_head(&root->log_writer_wait); |
1172 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1173 | init_waitqueue_head(&root->log_commit_wait[1]); | |
1174 | atomic_set(&root->log_commit[0], 0); | |
1175 | atomic_set(&root->log_commit[1], 0); | |
1176 | atomic_set(&root->log_writers, 0); | |
1177 | root->log_batch = 0; | |
1178 | root->log_transid = 0; | |
257c62e1 | 1179 | root->last_log_commit = 0; |
d0c803c4 | 1180 | extent_io_tree_init(&root->dirty_log_pages, |
f993c883 | 1181 | fs_info->btree_inode->i_mapping); |
017e5369 | 1182 | |
3768f368 CM |
1183 | memset(&root->root_key, 0, sizeof(root->root_key)); |
1184 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 1185 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 1186 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 1187 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 1188 | init_completion(&root->kobj_unregister); |
6702ed49 | 1189 | root->defrag_running = 0; |
4d775673 | 1190 | root->root_key.objectid = objectid; |
0ee5dc67 | 1191 | root->anon_dev = 0; |
3768f368 CM |
1192 | return 0; |
1193 | } | |
1194 | ||
db94535d | 1195 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
1196 | struct btrfs_fs_info *fs_info, |
1197 | u64 objectid, | |
e20d96d6 | 1198 | struct btrfs_root *root) |
3768f368 CM |
1199 | { |
1200 | int ret; | |
db94535d | 1201 | u32 blocksize; |
84234f3a | 1202 | u64 generation; |
3768f368 | 1203 | |
db94535d | 1204 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1205 | tree_root->sectorsize, tree_root->stripesize, |
1206 | root, fs_info, objectid); | |
3768f368 CM |
1207 | ret = btrfs_find_last_root(tree_root, objectid, |
1208 | &root->root_item, &root->root_key); | |
4df27c4d YZ |
1209 | if (ret > 0) |
1210 | return -ENOENT; | |
3768f368 CM |
1211 | BUG_ON(ret); |
1212 | ||
84234f3a | 1213 | generation = btrfs_root_generation(&root->root_item); |
db94535d | 1214 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
af31f5e5 | 1215 | root->commit_root = NULL; |
db94535d | 1216 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), |
84234f3a | 1217 | blocksize, generation); |
68433b73 CM |
1218 | if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) { |
1219 | free_extent_buffer(root->node); | |
af31f5e5 | 1220 | root->node = NULL; |
68433b73 CM |
1221 | return -EIO; |
1222 | } | |
4df27c4d | 1223 | root->commit_root = btrfs_root_node(root); |
d97e63b6 CM |
1224 | return 0; |
1225 | } | |
1226 | ||
7237f183 YZ |
1227 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1228 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1229 | { |
1230 | struct btrfs_root *root; | |
1231 | struct btrfs_root *tree_root = fs_info->tree_root; | |
7237f183 | 1232 | struct extent_buffer *leaf; |
e02119d5 CM |
1233 | |
1234 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
1235 | if (!root) | |
7237f183 | 1236 | return ERR_PTR(-ENOMEM); |
e02119d5 CM |
1237 | |
1238 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
1239 | tree_root->sectorsize, tree_root->stripesize, | |
1240 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1241 | ||
1242 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
1243 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1244 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
7237f183 YZ |
1245 | /* |
1246 | * log trees do not get reference counted because they go away | |
1247 | * before a real commit is actually done. They do store pointers | |
1248 | * to file data extents, and those reference counts still get | |
1249 | * updated (along with back refs to the log tree). | |
1250 | */ | |
e02119d5 CM |
1251 | root->ref_cows = 0; |
1252 | ||
5d4f98a2 | 1253 | leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, |
66d7e7f0 AJ |
1254 | BTRFS_TREE_LOG_OBJECTID, NULL, |
1255 | 0, 0, 0, 0); | |
7237f183 YZ |
1256 | if (IS_ERR(leaf)) { |
1257 | kfree(root); | |
1258 | return ERR_CAST(leaf); | |
1259 | } | |
e02119d5 | 1260 | |
5d4f98a2 YZ |
1261 | memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); |
1262 | btrfs_set_header_bytenr(leaf, leaf->start); | |
1263 | btrfs_set_header_generation(leaf, trans->transid); | |
1264 | btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); | |
1265 | btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); | |
7237f183 | 1266 | root->node = leaf; |
e02119d5 CM |
1267 | |
1268 | write_extent_buffer(root->node, root->fs_info->fsid, | |
1269 | (unsigned long)btrfs_header_fsid(root->node), | |
1270 | BTRFS_FSID_SIZE); | |
1271 | btrfs_mark_buffer_dirty(root->node); | |
1272 | btrfs_tree_unlock(root->node); | |
7237f183 YZ |
1273 | return root; |
1274 | } | |
1275 | ||
1276 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1277 | struct btrfs_fs_info *fs_info) | |
1278 | { | |
1279 | struct btrfs_root *log_root; | |
1280 | ||
1281 | log_root = alloc_log_tree(trans, fs_info); | |
1282 | if (IS_ERR(log_root)) | |
1283 | return PTR_ERR(log_root); | |
1284 | WARN_ON(fs_info->log_root_tree); | |
1285 | fs_info->log_root_tree = log_root; | |
1286 | return 0; | |
1287 | } | |
1288 | ||
1289 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1290 | struct btrfs_root *root) | |
1291 | { | |
1292 | struct btrfs_root *log_root; | |
1293 | struct btrfs_inode_item *inode_item; | |
1294 | ||
1295 | log_root = alloc_log_tree(trans, root->fs_info); | |
1296 | if (IS_ERR(log_root)) | |
1297 | return PTR_ERR(log_root); | |
1298 | ||
1299 | log_root->last_trans = trans->transid; | |
1300 | log_root->root_key.offset = root->root_key.objectid; | |
1301 | ||
1302 | inode_item = &log_root->root_item.inode; | |
1303 | inode_item->generation = cpu_to_le64(1); | |
1304 | inode_item->size = cpu_to_le64(3); | |
1305 | inode_item->nlink = cpu_to_le32(1); | |
1306 | inode_item->nbytes = cpu_to_le64(root->leafsize); | |
1307 | inode_item->mode = cpu_to_le32(S_IFDIR | 0755); | |
1308 | ||
5d4f98a2 | 1309 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1310 | |
1311 | WARN_ON(root->log_root); | |
1312 | root->log_root = log_root; | |
1313 | root->log_transid = 0; | |
257c62e1 | 1314 | root->last_log_commit = 0; |
e02119d5 CM |
1315 | return 0; |
1316 | } | |
1317 | ||
1318 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
1319 | struct btrfs_key *location) | |
1320 | { | |
1321 | struct btrfs_root *root; | |
1322 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 1323 | struct btrfs_path *path; |
5f39d397 | 1324 | struct extent_buffer *l; |
84234f3a | 1325 | u64 generation; |
db94535d | 1326 | u32 blocksize; |
0f7d52f4 CM |
1327 | int ret = 0; |
1328 | ||
5eda7b5e | 1329 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 1330 | if (!root) |
0f7d52f4 | 1331 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 1332 | if (location->offset == (u64)-1) { |
db94535d | 1333 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
1334 | location->objectid, root); |
1335 | if (ret) { | |
0f7d52f4 CM |
1336 | kfree(root); |
1337 | return ERR_PTR(ret); | |
1338 | } | |
13a8a7c8 | 1339 | goto out; |
0f7d52f4 CM |
1340 | } |
1341 | ||
db94535d | 1342 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1343 | tree_root->sectorsize, tree_root->stripesize, |
1344 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1345 | |
1346 | path = btrfs_alloc_path(); | |
db5b493a TI |
1347 | if (!path) { |
1348 | kfree(root); | |
1349 | return ERR_PTR(-ENOMEM); | |
1350 | } | |
0f7d52f4 | 1351 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); |
13a8a7c8 YZ |
1352 | if (ret == 0) { |
1353 | l = path->nodes[0]; | |
1354 | read_extent_buffer(l, &root->root_item, | |
1355 | btrfs_item_ptr_offset(l, path->slots[0]), | |
1356 | sizeof(root->root_item)); | |
1357 | memcpy(&root->root_key, location, sizeof(*location)); | |
0f7d52f4 | 1358 | } |
0f7d52f4 CM |
1359 | btrfs_free_path(path); |
1360 | if (ret) { | |
5e540f77 | 1361 | kfree(root); |
13a8a7c8 YZ |
1362 | if (ret > 0) |
1363 | ret = -ENOENT; | |
0f7d52f4 CM |
1364 | return ERR_PTR(ret); |
1365 | } | |
13a8a7c8 | 1366 | |
84234f3a | 1367 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1368 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1369 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1370 | blocksize, generation); |
5d4f98a2 | 1371 | root->commit_root = btrfs_root_node(root); |
0f7d52f4 | 1372 | BUG_ON(!root->node); |
13a8a7c8 | 1373 | out: |
08fe4db1 | 1374 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
e02119d5 | 1375 | root->ref_cows = 1; |
08fe4db1 LZ |
1376 | btrfs_check_and_init_root_item(&root->root_item); |
1377 | } | |
13a8a7c8 | 1378 | |
5eda7b5e CM |
1379 | return root; |
1380 | } | |
1381 | ||
edbd8d4e CM |
1382 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1383 | struct btrfs_key *location) | |
5eda7b5e CM |
1384 | { |
1385 | struct btrfs_root *root; | |
1386 | int ret; | |
1387 | ||
edbd8d4e CM |
1388 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1389 | return fs_info->tree_root; | |
1390 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1391 | return fs_info->extent_root; | |
8f18cf13 CM |
1392 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1393 | return fs_info->chunk_root; | |
1394 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1395 | return fs_info->dev_root; | |
0403e47e YZ |
1396 | if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) |
1397 | return fs_info->csum_root; | |
4df27c4d YZ |
1398 | again: |
1399 | spin_lock(&fs_info->fs_roots_radix_lock); | |
5eda7b5e CM |
1400 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1401 | (unsigned long)location->objectid); | |
4df27c4d | 1402 | spin_unlock(&fs_info->fs_roots_radix_lock); |
5eda7b5e CM |
1403 | if (root) |
1404 | return root; | |
1405 | ||
e02119d5 | 1406 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1407 | if (IS_ERR(root)) |
1408 | return root; | |
3394e160 | 1409 | |
581bb050 | 1410 | root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); |
581bb050 LZ |
1411 | root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), |
1412 | GFP_NOFS); | |
35a30d7c DS |
1413 | if (!root->free_ino_pinned || !root->free_ino_ctl) { |
1414 | ret = -ENOMEM; | |
581bb050 | 1415 | goto fail; |
35a30d7c | 1416 | } |
581bb050 LZ |
1417 | |
1418 | btrfs_init_free_ino_ctl(root); | |
1419 | mutex_init(&root->fs_commit_mutex); | |
1420 | spin_lock_init(&root->cache_lock); | |
1421 | init_waitqueue_head(&root->cache_wait); | |
1422 | ||
0ee5dc67 | 1423 | ret = get_anon_bdev(&root->anon_dev); |
ac08aedf CM |
1424 | if (ret) |
1425 | goto fail; | |
3394e160 | 1426 | |
d68fc57b YZ |
1427 | if (btrfs_root_refs(&root->root_item) == 0) { |
1428 | ret = -ENOENT; | |
1429 | goto fail; | |
1430 | } | |
1431 | ||
1432 | ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid); | |
1433 | if (ret < 0) | |
1434 | goto fail; | |
1435 | if (ret == 0) | |
1436 | root->orphan_item_inserted = 1; | |
1437 | ||
4df27c4d YZ |
1438 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
1439 | if (ret) | |
1440 | goto fail; | |
1441 | ||
1442 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2619ba1f CM |
1443 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1444 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 | 1445 | root); |
d68fc57b | 1446 | if (ret == 0) |
4df27c4d | 1447 | root->in_radix = 1; |
d68fc57b | 1448 | |
4df27c4d YZ |
1449 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1450 | radix_tree_preload_end(); | |
0f7d52f4 | 1451 | if (ret) { |
4df27c4d YZ |
1452 | if (ret == -EEXIST) { |
1453 | free_fs_root(root); | |
1454 | goto again; | |
1455 | } | |
1456 | goto fail; | |
0f7d52f4 | 1457 | } |
4df27c4d YZ |
1458 | |
1459 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1460 | root->root_key.objectid); | |
1461 | WARN_ON(ret); | |
edbd8d4e | 1462 | return root; |
4df27c4d YZ |
1463 | fail: |
1464 | free_fs_root(root); | |
1465 | return ERR_PTR(ret); | |
edbd8d4e CM |
1466 | } |
1467 | ||
04160088 CM |
1468 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) |
1469 | { | |
1470 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1471 | int ret = 0; | |
04160088 CM |
1472 | struct btrfs_device *device; |
1473 | struct backing_dev_info *bdi; | |
b7967db7 | 1474 | |
1f78160c XG |
1475 | rcu_read_lock(); |
1476 | list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { | |
dfe25020 CM |
1477 | if (!device->bdev) |
1478 | continue; | |
04160088 CM |
1479 | bdi = blk_get_backing_dev_info(device->bdev); |
1480 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1481 | ret = 1; | |
1482 | break; | |
1483 | } | |
1484 | } | |
1f78160c | 1485 | rcu_read_unlock(); |
04160088 CM |
1486 | return ret; |
1487 | } | |
1488 | ||
ad081f14 JA |
1489 | /* |
1490 | * If this fails, caller must call bdi_destroy() to get rid of the | |
1491 | * bdi again. | |
1492 | */ | |
04160088 CM |
1493 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) |
1494 | { | |
ad081f14 JA |
1495 | int err; |
1496 | ||
1497 | bdi->capabilities = BDI_CAP_MAP_COPY; | |
e6d086d8 | 1498 | err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); |
ad081f14 JA |
1499 | if (err) |
1500 | return err; | |
1501 | ||
4575c9cc | 1502 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1503 | bdi->congested_fn = btrfs_congested_fn; |
1504 | bdi->congested_data = info; | |
1505 | return 0; | |
1506 | } | |
1507 | ||
ce9adaa5 CM |
1508 | static int bio_ready_for_csum(struct bio *bio) |
1509 | { | |
1510 | u64 length = 0; | |
1511 | u64 buf_len = 0; | |
1512 | u64 start = 0; | |
1513 | struct page *page; | |
1514 | struct extent_io_tree *io_tree = NULL; | |
ce9adaa5 CM |
1515 | struct bio_vec *bvec; |
1516 | int i; | |
1517 | int ret; | |
1518 | ||
1519 | bio_for_each_segment(bvec, bio, i) { | |
1520 | page = bvec->bv_page; | |
1521 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1522 | length += bvec->bv_len; | |
1523 | continue; | |
1524 | } | |
1525 | if (!page->private) { | |
1526 | length += bvec->bv_len; | |
1527 | continue; | |
1528 | } | |
1529 | length = bvec->bv_len; | |
1530 | buf_len = page->private >> 2; | |
1531 | start = page_offset(page) + bvec->bv_offset; | |
1532 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
ce9adaa5 CM |
1533 | } |
1534 | /* are we fully contained in this bio? */ | |
1535 | if (buf_len <= length) | |
1536 | return 1; | |
1537 | ||
1538 | ret = extent_range_uptodate(io_tree, start + length, | |
1539 | start + buf_len - 1); | |
ce9adaa5 CM |
1540 | return ret; |
1541 | } | |
1542 | ||
8b712842 CM |
1543 | /* |
1544 | * called by the kthread helper functions to finally call the bio end_io | |
1545 | * functions. This is where read checksum verification actually happens | |
1546 | */ | |
1547 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1548 | { |
ce9adaa5 | 1549 | struct bio *bio; |
8b712842 CM |
1550 | struct end_io_wq *end_io_wq; |
1551 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1552 | int error; |
ce9adaa5 | 1553 | |
8b712842 CM |
1554 | end_io_wq = container_of(work, struct end_io_wq, work); |
1555 | bio = end_io_wq->bio; | |
1556 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1557 | |
cad321ad | 1558 | /* metadata bio reads are special because the whole tree block must |
8b712842 CM |
1559 | * be checksummed at once. This makes sure the entire block is in |
1560 | * ram and up to date before trying to verify things. For | |
1561 | * blocksize <= pagesize, it is basically a noop | |
1562 | */ | |
7b6d91da | 1563 | if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata && |
cad321ad | 1564 | !bio_ready_for_csum(bio)) { |
d20f7043 | 1565 | btrfs_queue_worker(&fs_info->endio_meta_workers, |
8b712842 CM |
1566 | &end_io_wq->work); |
1567 | return; | |
1568 | } | |
1569 | error = end_io_wq->error; | |
1570 | bio->bi_private = end_io_wq->private; | |
1571 | bio->bi_end_io = end_io_wq->end_io; | |
1572 | kfree(end_io_wq); | |
8b712842 | 1573 | bio_endio(bio, error); |
44b8bd7e CM |
1574 | } |
1575 | ||
a74a4b97 CM |
1576 | static int cleaner_kthread(void *arg) |
1577 | { | |
1578 | struct btrfs_root *root = arg; | |
1579 | ||
1580 | do { | |
a74a4b97 | 1581 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); |
76dda93c YZ |
1582 | |
1583 | if (!(root->fs_info->sb->s_flags & MS_RDONLY) && | |
1584 | mutex_trylock(&root->fs_info->cleaner_mutex)) { | |
24bbcf04 | 1585 | btrfs_run_delayed_iputs(root); |
76dda93c YZ |
1586 | btrfs_clean_old_snapshots(root); |
1587 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
4cb5300b | 1588 | btrfs_run_defrag_inodes(root->fs_info); |
76dda93c | 1589 | } |
a74a4b97 CM |
1590 | |
1591 | if (freezing(current)) { | |
1592 | refrigerator(); | |
1593 | } else { | |
a74a4b97 | 1594 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1595 | if (!kthread_should_stop()) |
1596 | schedule(); | |
a74a4b97 CM |
1597 | __set_current_state(TASK_RUNNING); |
1598 | } | |
1599 | } while (!kthread_should_stop()); | |
1600 | return 0; | |
1601 | } | |
1602 | ||
1603 | static int transaction_kthread(void *arg) | |
1604 | { | |
1605 | struct btrfs_root *root = arg; | |
1606 | struct btrfs_trans_handle *trans; | |
1607 | struct btrfs_transaction *cur; | |
8929ecfa | 1608 | u64 transid; |
a74a4b97 CM |
1609 | unsigned long now; |
1610 | unsigned long delay; | |
1611 | int ret; | |
1612 | ||
1613 | do { | |
a74a4b97 CM |
1614 | delay = HZ * 30; |
1615 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1616 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1617 | ||
a4abeea4 | 1618 | spin_lock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1619 | cur = root->fs_info->running_transaction; |
1620 | if (!cur) { | |
a4abeea4 | 1621 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1622 | goto sleep; |
1623 | } | |
31153d81 | 1624 | |
a74a4b97 | 1625 | now = get_seconds(); |
8929ecfa YZ |
1626 | if (!cur->blocked && |
1627 | (now < cur->start_time || now - cur->start_time < 30)) { | |
a4abeea4 | 1628 | spin_unlock(&root->fs_info->trans_lock); |
a74a4b97 CM |
1629 | delay = HZ * 5; |
1630 | goto sleep; | |
1631 | } | |
8929ecfa | 1632 | transid = cur->transid; |
a4abeea4 | 1633 | spin_unlock(&root->fs_info->trans_lock); |
56bec294 | 1634 | |
7a7eaa40 | 1635 | trans = btrfs_join_transaction(root); |
3612b495 | 1636 | BUG_ON(IS_ERR(trans)); |
8929ecfa YZ |
1637 | if (transid == trans->transid) { |
1638 | ret = btrfs_commit_transaction(trans, root); | |
1639 | BUG_ON(ret); | |
1640 | } else { | |
1641 | btrfs_end_transaction(trans, root); | |
1642 | } | |
a74a4b97 CM |
1643 | sleep: |
1644 | wake_up_process(root->fs_info->cleaner_kthread); | |
1645 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1646 | ||
1647 | if (freezing(current)) { | |
1648 | refrigerator(); | |
1649 | } else { | |
a74a4b97 | 1650 | set_current_state(TASK_INTERRUPTIBLE); |
8929ecfa YZ |
1651 | if (!kthread_should_stop() && |
1652 | !btrfs_transaction_blocked(root->fs_info)) | |
1653 | schedule_timeout(delay); | |
a74a4b97 CM |
1654 | __set_current_state(TASK_RUNNING); |
1655 | } | |
1656 | } while (!kthread_should_stop()); | |
1657 | return 0; | |
1658 | } | |
1659 | ||
af31f5e5 CM |
1660 | /* |
1661 | * this will find the highest generation in the array of | |
1662 | * root backups. The index of the highest array is returned, | |
1663 | * or -1 if we can't find anything. | |
1664 | * | |
1665 | * We check to make sure the array is valid by comparing the | |
1666 | * generation of the latest root in the array with the generation | |
1667 | * in the super block. If they don't match we pitch it. | |
1668 | */ | |
1669 | static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) | |
1670 | { | |
1671 | u64 cur; | |
1672 | int newest_index = -1; | |
1673 | struct btrfs_root_backup *root_backup; | |
1674 | int i; | |
1675 | ||
1676 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1677 | root_backup = info->super_copy->super_roots + i; | |
1678 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1679 | if (cur == newest_gen) | |
1680 | newest_index = i; | |
1681 | } | |
1682 | ||
1683 | /* check to see if we actually wrapped around */ | |
1684 | if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { | |
1685 | root_backup = info->super_copy->super_roots; | |
1686 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1687 | if (cur == newest_gen) | |
1688 | newest_index = 0; | |
1689 | } | |
1690 | return newest_index; | |
1691 | } | |
1692 | ||
1693 | ||
1694 | /* | |
1695 | * find the oldest backup so we know where to store new entries | |
1696 | * in the backup array. This will set the backup_root_index | |
1697 | * field in the fs_info struct | |
1698 | */ | |
1699 | static void find_oldest_super_backup(struct btrfs_fs_info *info, | |
1700 | u64 newest_gen) | |
1701 | { | |
1702 | int newest_index = -1; | |
1703 | ||
1704 | newest_index = find_newest_super_backup(info, newest_gen); | |
1705 | /* if there was garbage in there, just move along */ | |
1706 | if (newest_index == -1) { | |
1707 | info->backup_root_index = 0; | |
1708 | } else { | |
1709 | info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1710 | } | |
1711 | } | |
1712 | ||
1713 | /* | |
1714 | * copy all the root pointers into the super backup array. | |
1715 | * this will bump the backup pointer by one when it is | |
1716 | * done | |
1717 | */ | |
1718 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1719 | { | |
1720 | int next_backup; | |
1721 | struct btrfs_root_backup *root_backup; | |
1722 | int last_backup; | |
1723 | ||
1724 | next_backup = info->backup_root_index; | |
1725 | last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1726 | BTRFS_NUM_BACKUP_ROOTS; | |
1727 | ||
1728 | /* | |
1729 | * just overwrite the last backup if we're at the same generation | |
1730 | * this happens only at umount | |
1731 | */ | |
1732 | root_backup = info->super_for_commit->super_roots + last_backup; | |
1733 | if (btrfs_backup_tree_root_gen(root_backup) == | |
1734 | btrfs_header_generation(info->tree_root->node)) | |
1735 | next_backup = last_backup; | |
1736 | ||
1737 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1738 | ||
1739 | /* | |
1740 | * make sure all of our padding and empty slots get zero filled | |
1741 | * regardless of which ones we use today | |
1742 | */ | |
1743 | memset(root_backup, 0, sizeof(*root_backup)); | |
1744 | ||
1745 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1746 | ||
1747 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1748 | btrfs_set_backup_tree_root_gen(root_backup, | |
1749 | btrfs_header_generation(info->tree_root->node)); | |
1750 | ||
1751 | btrfs_set_backup_tree_root_level(root_backup, | |
1752 | btrfs_header_level(info->tree_root->node)); | |
1753 | ||
1754 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1755 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1756 | btrfs_header_generation(info->chunk_root->node)); | |
1757 | btrfs_set_backup_chunk_root_level(root_backup, | |
1758 | btrfs_header_level(info->chunk_root->node)); | |
1759 | ||
1760 | btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); | |
1761 | btrfs_set_backup_extent_root_gen(root_backup, | |
1762 | btrfs_header_generation(info->extent_root->node)); | |
1763 | btrfs_set_backup_extent_root_level(root_backup, | |
1764 | btrfs_header_level(info->extent_root->node)); | |
1765 | ||
7c7e82a7 CM |
1766 | /* |
1767 | * we might commit during log recovery, which happens before we set | |
1768 | * the fs_root. Make sure it is valid before we fill it in. | |
1769 | */ | |
1770 | if (info->fs_root && info->fs_root->node) { | |
1771 | btrfs_set_backup_fs_root(root_backup, | |
1772 | info->fs_root->node->start); | |
1773 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1774 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1775 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1776 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1777 | } |
af31f5e5 CM |
1778 | |
1779 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1780 | btrfs_set_backup_dev_root_gen(root_backup, | |
1781 | btrfs_header_generation(info->dev_root->node)); | |
1782 | btrfs_set_backup_dev_root_level(root_backup, | |
1783 | btrfs_header_level(info->dev_root->node)); | |
1784 | ||
1785 | btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); | |
1786 | btrfs_set_backup_csum_root_gen(root_backup, | |
1787 | btrfs_header_generation(info->csum_root->node)); | |
1788 | btrfs_set_backup_csum_root_level(root_backup, | |
1789 | btrfs_header_level(info->csum_root->node)); | |
1790 | ||
1791 | btrfs_set_backup_total_bytes(root_backup, | |
1792 | btrfs_super_total_bytes(info->super_copy)); | |
1793 | btrfs_set_backup_bytes_used(root_backup, | |
1794 | btrfs_super_bytes_used(info->super_copy)); | |
1795 | btrfs_set_backup_num_devices(root_backup, | |
1796 | btrfs_super_num_devices(info->super_copy)); | |
1797 | ||
1798 | /* | |
1799 | * if we don't copy this out to the super_copy, it won't get remembered | |
1800 | * for the next commit | |
1801 | */ | |
1802 | memcpy(&info->super_copy->super_roots, | |
1803 | &info->super_for_commit->super_roots, | |
1804 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1805 | } | |
1806 | ||
1807 | /* | |
1808 | * this copies info out of the root backup array and back into | |
1809 | * the in-memory super block. It is meant to help iterate through | |
1810 | * the array, so you send it the number of backups you've already | |
1811 | * tried and the last backup index you used. | |
1812 | * | |
1813 | * this returns -1 when it has tried all the backups | |
1814 | */ | |
1815 | static noinline int next_root_backup(struct btrfs_fs_info *info, | |
1816 | struct btrfs_super_block *super, | |
1817 | int *num_backups_tried, int *backup_index) | |
1818 | { | |
1819 | struct btrfs_root_backup *root_backup; | |
1820 | int newest = *backup_index; | |
1821 | ||
1822 | if (*num_backups_tried == 0) { | |
1823 | u64 gen = btrfs_super_generation(super); | |
1824 | ||
1825 | newest = find_newest_super_backup(info, gen); | |
1826 | if (newest == -1) | |
1827 | return -1; | |
1828 | ||
1829 | *backup_index = newest; | |
1830 | *num_backups_tried = 1; | |
1831 | } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { | |
1832 | /* we've tried all the backups, all done */ | |
1833 | return -1; | |
1834 | } else { | |
1835 | /* jump to the next oldest backup */ | |
1836 | newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % | |
1837 | BTRFS_NUM_BACKUP_ROOTS; | |
1838 | *backup_index = newest; | |
1839 | *num_backups_tried += 1; | |
1840 | } | |
1841 | root_backup = super->super_roots + newest; | |
1842 | ||
1843 | btrfs_set_super_generation(super, | |
1844 | btrfs_backup_tree_root_gen(root_backup)); | |
1845 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1846 | btrfs_set_super_root_level(super, | |
1847 | btrfs_backup_tree_root_level(root_backup)); | |
1848 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1849 | ||
1850 | /* | |
1851 | * fixme: the total bytes and num_devices need to match or we should | |
1852 | * need a fsck | |
1853 | */ | |
1854 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1855 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1856 | return 0; | |
1857 | } | |
1858 | ||
1859 | /* helper to cleanup tree roots */ | |
1860 | static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) | |
1861 | { | |
1862 | free_extent_buffer(info->tree_root->node); | |
1863 | free_extent_buffer(info->tree_root->commit_root); | |
1864 | free_extent_buffer(info->dev_root->node); | |
1865 | free_extent_buffer(info->dev_root->commit_root); | |
1866 | free_extent_buffer(info->extent_root->node); | |
1867 | free_extent_buffer(info->extent_root->commit_root); | |
1868 | free_extent_buffer(info->csum_root->node); | |
1869 | free_extent_buffer(info->csum_root->commit_root); | |
1870 | ||
1871 | info->tree_root->node = NULL; | |
1872 | info->tree_root->commit_root = NULL; | |
1873 | info->dev_root->node = NULL; | |
1874 | info->dev_root->commit_root = NULL; | |
1875 | info->extent_root->node = NULL; | |
1876 | info->extent_root->commit_root = NULL; | |
1877 | info->csum_root->node = NULL; | |
1878 | info->csum_root->commit_root = NULL; | |
1879 | ||
1880 | if (chunk_root) { | |
1881 | free_extent_buffer(info->chunk_root->node); | |
1882 | free_extent_buffer(info->chunk_root->commit_root); | |
1883 | info->chunk_root->node = NULL; | |
1884 | info->chunk_root->commit_root = NULL; | |
1885 | } | |
1886 | } | |
1887 | ||
1888 | ||
8a4b83cc | 1889 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1890 | struct btrfs_fs_devices *fs_devices, |
1891 | char *options) | |
2e635a27 | 1892 | { |
db94535d CM |
1893 | u32 sectorsize; |
1894 | u32 nodesize; | |
1895 | u32 leafsize; | |
1896 | u32 blocksize; | |
87ee04eb | 1897 | u32 stripesize; |
84234f3a | 1898 | u64 generation; |
f2b636e8 | 1899 | u64 features; |
3de4586c | 1900 | struct btrfs_key location; |
a061fc8d | 1901 | struct buffer_head *bh; |
4d34b278 | 1902 | struct btrfs_super_block *disk_super; |
450ba0ea | 1903 | struct btrfs_root *tree_root = btrfs_sb(sb); |
4d34b278 ID |
1904 | struct btrfs_fs_info *fs_info = tree_root->fs_info; |
1905 | struct btrfs_root *extent_root; | |
1906 | struct btrfs_root *csum_root; | |
1907 | struct btrfs_root *chunk_root; | |
1908 | struct btrfs_root *dev_root; | |
e02119d5 | 1909 | struct btrfs_root *log_tree_root; |
eb60ceac | 1910 | int ret; |
e58ca020 | 1911 | int err = -EINVAL; |
af31f5e5 CM |
1912 | int num_backups_tried = 0; |
1913 | int backup_index = 0; | |
4543df7e | 1914 | |
4d34b278 ID |
1915 | extent_root = fs_info->extent_root = |
1916 | kzalloc(sizeof(struct btrfs_root), GFP_NOFS); | |
1917 | csum_root = fs_info->csum_root = | |
1918 | kzalloc(sizeof(struct btrfs_root), GFP_NOFS); | |
1919 | chunk_root = fs_info->chunk_root = | |
1920 | kzalloc(sizeof(struct btrfs_root), GFP_NOFS); | |
1921 | dev_root = fs_info->dev_root = | |
1922 | kzalloc(sizeof(struct btrfs_root), GFP_NOFS); | |
8790d502 | 1923 | |
4d34b278 | 1924 | if (!extent_root || !csum_root || !chunk_root || !dev_root) { |
39279cc3 CM |
1925 | err = -ENOMEM; |
1926 | goto fail; | |
1927 | } | |
76dda93c YZ |
1928 | |
1929 | ret = init_srcu_struct(&fs_info->subvol_srcu); | |
1930 | if (ret) { | |
1931 | err = ret; | |
1932 | goto fail; | |
1933 | } | |
1934 | ||
1935 | ret = setup_bdi(fs_info, &fs_info->bdi); | |
1936 | if (ret) { | |
1937 | err = ret; | |
1938 | goto fail_srcu; | |
1939 | } | |
1940 | ||
1941 | fs_info->btree_inode = new_inode(sb); | |
1942 | if (!fs_info->btree_inode) { | |
1943 | err = -ENOMEM; | |
1944 | goto fail_bdi; | |
1945 | } | |
1946 | ||
a6591715 | 1947 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
1561deda | 1948 | |
76dda93c | 1949 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
8fd17795 | 1950 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1951 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 1952 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
19c00ddc | 1953 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1954 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
5a3f23d5 | 1955 | INIT_LIST_HEAD(&fs_info->ordered_operations); |
11833d66 | 1956 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
1832a6d5 | 1957 | spin_lock_init(&fs_info->delalloc_lock); |
a4abeea4 | 1958 | spin_lock_init(&fs_info->trans_lock); |
31153d81 | 1959 | spin_lock_init(&fs_info->ref_cache_lock); |
76dda93c | 1960 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 1961 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 1962 | spin_lock_init(&fs_info->defrag_inodes_lock); |
2bf64758 | 1963 | spin_lock_init(&fs_info->free_chunk_lock); |
7585717f | 1964 | mutex_init(&fs_info->reloc_mutex); |
19c00ddc | 1965 | |
58176a96 | 1966 | init_completion(&fs_info->kobj_unregister); |
0b86a832 | 1967 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1968 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1969 | btrfs_mapping_init(&fs_info->mapping_tree); |
f0486c68 YZ |
1970 | btrfs_init_block_rsv(&fs_info->global_block_rsv); |
1971 | btrfs_init_block_rsv(&fs_info->delalloc_block_rsv); | |
1972 | btrfs_init_block_rsv(&fs_info->trans_block_rsv); | |
1973 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv); | |
1974 | btrfs_init_block_rsv(&fs_info->empty_block_rsv); | |
6d668dda | 1975 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv); |
cb03c743 | 1976 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1977 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1978 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1979 | atomic_set(&fs_info->nr_async_bios, 0); |
4cb5300b | 1980 | atomic_set(&fs_info->defrag_running, 0); |
e20d96d6 | 1981 | fs_info->sb = sb; |
6f568d35 | 1982 | fs_info->max_inline = 8192 * 1024; |
9ed74f2d | 1983 | fs_info->metadata_ratio = 0; |
4cb5300b | 1984 | fs_info->defrag_inodes = RB_ROOT; |
a4abeea4 | 1985 | fs_info->trans_no_join = 0; |
2bf64758 | 1986 | fs_info->free_chunk_space = 0; |
c8b97818 | 1987 | |
90519d66 AJ |
1988 | /* readahead state */ |
1989 | INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); | |
1990 | spin_lock_init(&fs_info->reada_lock); | |
c8b97818 | 1991 | |
b34b086c CM |
1992 | fs_info->thread_pool_size = min_t(unsigned long, |
1993 | num_online_cpus() + 2, 8); | |
0afbaf8c | 1994 | |
3eaa2885 CM |
1995 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1996 | spin_lock_init(&fs_info->ordered_extent_lock); | |
16cdcec7 MX |
1997 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), |
1998 | GFP_NOFS); | |
1999 | if (!fs_info->delayed_root) { | |
2000 | err = -ENOMEM; | |
2001 | goto fail_iput; | |
2002 | } | |
2003 | btrfs_init_delayed_root(fs_info->delayed_root); | |
3eaa2885 | 2004 | |
a2de733c AJ |
2005 | mutex_init(&fs_info->scrub_lock); |
2006 | atomic_set(&fs_info->scrubs_running, 0); | |
2007 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2008 | atomic_set(&fs_info->scrubs_paused, 0); | |
2009 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2010 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
2011 | init_rwsem(&fs_info->scrub_super_lock); | |
2012 | fs_info->scrub_workers_refcnt = 0; | |
21adbd5c SB |
2013 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2014 | fs_info->check_integrity_print_mask = 0; | |
2015 | #endif | |
a2de733c | 2016 | |
c9e9f97b ID |
2017 | spin_lock_init(&fs_info->balance_lock); |
2018 | mutex_init(&fs_info->balance_mutex); | |
837d5b6e ID |
2019 | atomic_set(&fs_info->balance_running, 0); |
2020 | atomic_set(&fs_info->balance_pause_req, 0); | |
a7e99c69 | 2021 | atomic_set(&fs_info->balance_cancel_req, 0); |
c9e9f97b | 2022 | fs_info->balance_ctl = NULL; |
837d5b6e | 2023 | init_waitqueue_head(&fs_info->balance_wait_q); |
c9e9f97b | 2024 | |
a061fc8d CM |
2025 | sb->s_blocksize = 4096; |
2026 | sb->s_blocksize_bits = blksize_bits(4096); | |
32a88aa1 | 2027 | sb->s_bdi = &fs_info->bdi; |
a061fc8d | 2028 | |
76dda93c | 2029 | fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; |
bfe86848 | 2030 | set_nlink(fs_info->btree_inode, 1); |
0afbaf8c CM |
2031 | /* |
2032 | * we set the i_size on the btree inode to the max possible int. | |
2033 | * the real end of the address space is determined by all of | |
2034 | * the devices in the system | |
2035 | */ | |
2036 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 2037 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
2038 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
2039 | ||
5d4f98a2 | 2040 | RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); |
d1310b2e | 2041 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
f993c883 | 2042 | fs_info->btree_inode->i_mapping); |
a8067e02 | 2043 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); |
d1310b2e CM |
2044 | |
2045 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 2046 | |
76dda93c YZ |
2047 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
2048 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
2049 | sizeof(struct btrfs_key)); | |
2050 | BTRFS_I(fs_info->btree_inode)->dummy_inode = 1; | |
c65ddb52 | 2051 | insert_inode_hash(fs_info->btree_inode); |
76dda93c | 2052 | |
0f9dd46c | 2053 | spin_lock_init(&fs_info->block_group_cache_lock); |
6bef4d31 | 2054 | fs_info->block_group_cache_tree = RB_ROOT; |
0f9dd46c | 2055 | |
11833d66 | 2056 | extent_io_tree_init(&fs_info->freed_extents[0], |
f993c883 | 2057 | fs_info->btree_inode->i_mapping); |
11833d66 | 2058 | extent_io_tree_init(&fs_info->freed_extents[1], |
f993c883 | 2059 | fs_info->btree_inode->i_mapping); |
11833d66 | 2060 | fs_info->pinned_extents = &fs_info->freed_extents[0]; |
e66f709b | 2061 | fs_info->do_barriers = 1; |
e18e4809 | 2062 | |
39279cc3 | 2063 | |
5a3f23d5 | 2064 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2065 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2066 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2067 | mutex_init(&fs_info->transaction_kthread_mutex); |
2068 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 2069 | mutex_init(&fs_info->volume_mutex); |
276e680d | 2070 | init_rwsem(&fs_info->extent_commit_sem); |
c71bf099 | 2071 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2072 | init_rwsem(&fs_info->subvol_sem); |
fa9c0d79 CM |
2073 | |
2074 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); | |
2075 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2076 | ||
e6dcd2dc | 2077 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2078 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2079 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2080 | init_waitqueue_head(&fs_info->async_submit_wait); |
3768f368 | 2081 | |
0b86a832 | 2082 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 2083 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 2084 | |
a512bbf8 | 2085 | bh = btrfs_read_dev_super(fs_devices->latest_bdev); |
20b45077 DY |
2086 | if (!bh) { |
2087 | err = -EINVAL; | |
16cdcec7 | 2088 | goto fail_alloc; |
20b45077 | 2089 | } |
39279cc3 | 2090 | |
6c41761f DS |
2091 | memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); |
2092 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
2093 | sizeof(*fs_info->super_for_commit)); | |
a061fc8d | 2094 | brelse(bh); |
5f39d397 | 2095 | |
6c41761f | 2096 | memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); |
0b86a832 | 2097 | |
6c41761f | 2098 | disk_super = fs_info->super_copy; |
0f7d52f4 | 2099 | if (!btrfs_super_root(disk_super)) |
16cdcec7 | 2100 | goto fail_alloc; |
0f7d52f4 | 2101 | |
acce952b | 2102 | /* check FS state, whether FS is broken. */ |
2103 | fs_info->fs_state |= btrfs_super_flags(disk_super); | |
2104 | ||
2105 | btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); | |
2106 | ||
af31f5e5 CM |
2107 | /* |
2108 | * run through our array of backup supers and setup | |
2109 | * our ring pointer to the oldest one | |
2110 | */ | |
2111 | generation = btrfs_super_generation(disk_super); | |
2112 | find_oldest_super_backup(fs_info, generation); | |
2113 | ||
75e7cb7f LB |
2114 | /* |
2115 | * In the long term, we'll store the compression type in the super | |
2116 | * block, and it'll be used for per file compression control. | |
2117 | */ | |
2118 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
2119 | ||
2b82032c YZ |
2120 | ret = btrfs_parse_options(tree_root, options); |
2121 | if (ret) { | |
2122 | err = ret; | |
16cdcec7 | 2123 | goto fail_alloc; |
2b82032c | 2124 | } |
dfe25020 | 2125 | |
f2b636e8 JB |
2126 | features = btrfs_super_incompat_flags(disk_super) & |
2127 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
2128 | if (features) { | |
2129 | printk(KERN_ERR "BTRFS: couldn't mount because of " | |
2130 | "unsupported optional features (%Lx).\n", | |
21380931 | 2131 | (unsigned long long)features); |
f2b636e8 | 2132 | err = -EINVAL; |
16cdcec7 | 2133 | goto fail_alloc; |
f2b636e8 JB |
2134 | } |
2135 | ||
5d4f98a2 | 2136 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae LZ |
2137 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
2138 | if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO) | |
2139 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
2140 | btrfs_set_super_incompat_flags(disk_super, features); | |
5d4f98a2 | 2141 | |
f2b636e8 JB |
2142 | features = btrfs_super_compat_ro_flags(disk_super) & |
2143 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
2144 | if (!(sb->s_flags & MS_RDONLY) && features) { | |
2145 | printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " | |
2146 | "unsupported option features (%Lx).\n", | |
21380931 | 2147 | (unsigned long long)features); |
f2b636e8 | 2148 | err = -EINVAL; |
16cdcec7 | 2149 | goto fail_alloc; |
f2b636e8 | 2150 | } |
61d92c32 CM |
2151 | |
2152 | btrfs_init_workers(&fs_info->generic_worker, | |
2153 | "genwork", 1, NULL); | |
2154 | ||
5443be45 | 2155 | btrfs_init_workers(&fs_info->workers, "worker", |
61d92c32 CM |
2156 | fs_info->thread_pool_size, |
2157 | &fs_info->generic_worker); | |
c8b97818 | 2158 | |
771ed689 | 2159 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
61d92c32 CM |
2160 | fs_info->thread_pool_size, |
2161 | &fs_info->generic_worker); | |
771ed689 | 2162 | |
5443be45 | 2163 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 | 2164 | min_t(u64, fs_devices->num_devices, |
61d92c32 CM |
2165 | fs_info->thread_pool_size), |
2166 | &fs_info->generic_worker); | |
61b49440 | 2167 | |
bab39bf9 JB |
2168 | btrfs_init_workers(&fs_info->caching_workers, "cache", |
2169 | 2, &fs_info->generic_worker); | |
2170 | ||
61b49440 CM |
2171 | /* a higher idle thresh on the submit workers makes it much more |
2172 | * likely that bios will be send down in a sane order to the | |
2173 | * devices | |
2174 | */ | |
2175 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 2176 | |
771ed689 | 2177 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 2178 | fs_info->workers.ordered = 1; |
61b49440 | 2179 | |
771ed689 CM |
2180 | fs_info->delalloc_workers.idle_thresh = 2; |
2181 | fs_info->delalloc_workers.ordered = 1; | |
2182 | ||
61d92c32 CM |
2183 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, |
2184 | &fs_info->generic_worker); | |
5443be45 | 2185 | btrfs_init_workers(&fs_info->endio_workers, "endio", |
61d92c32 CM |
2186 | fs_info->thread_pool_size, |
2187 | &fs_info->generic_worker); | |
d20f7043 | 2188 | btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", |
61d92c32 CM |
2189 | fs_info->thread_pool_size, |
2190 | &fs_info->generic_worker); | |
cad321ad | 2191 | btrfs_init_workers(&fs_info->endio_meta_write_workers, |
61d92c32 CM |
2192 | "endio-meta-write", fs_info->thread_pool_size, |
2193 | &fs_info->generic_worker); | |
5443be45 | 2194 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", |
61d92c32 CM |
2195 | fs_info->thread_pool_size, |
2196 | &fs_info->generic_worker); | |
0cb59c99 JB |
2197 | btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", |
2198 | 1, &fs_info->generic_worker); | |
16cdcec7 MX |
2199 | btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", |
2200 | fs_info->thread_pool_size, | |
2201 | &fs_info->generic_worker); | |
90519d66 AJ |
2202 | btrfs_init_workers(&fs_info->readahead_workers, "readahead", |
2203 | fs_info->thread_pool_size, | |
2204 | &fs_info->generic_worker); | |
61b49440 CM |
2205 | |
2206 | /* | |
2207 | * endios are largely parallel and should have a very | |
2208 | * low idle thresh | |
2209 | */ | |
2210 | fs_info->endio_workers.idle_thresh = 4; | |
b51912c9 CM |
2211 | fs_info->endio_meta_workers.idle_thresh = 4; |
2212 | ||
9042846b CM |
2213 | fs_info->endio_write_workers.idle_thresh = 2; |
2214 | fs_info->endio_meta_write_workers.idle_thresh = 2; | |
90519d66 | 2215 | fs_info->readahead_workers.idle_thresh = 2; |
9042846b | 2216 | |
0dc3b84a JB |
2217 | /* |
2218 | * btrfs_start_workers can really only fail because of ENOMEM so just | |
2219 | * return -ENOMEM if any of these fail. | |
2220 | */ | |
2221 | ret = btrfs_start_workers(&fs_info->workers); | |
2222 | ret |= btrfs_start_workers(&fs_info->generic_worker); | |
2223 | ret |= btrfs_start_workers(&fs_info->submit_workers); | |
2224 | ret |= btrfs_start_workers(&fs_info->delalloc_workers); | |
2225 | ret |= btrfs_start_workers(&fs_info->fixup_workers); | |
2226 | ret |= btrfs_start_workers(&fs_info->endio_workers); | |
2227 | ret |= btrfs_start_workers(&fs_info->endio_meta_workers); | |
2228 | ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); | |
2229 | ret |= btrfs_start_workers(&fs_info->endio_write_workers); | |
2230 | ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); | |
2231 | ret |= btrfs_start_workers(&fs_info->delayed_workers); | |
2232 | ret |= btrfs_start_workers(&fs_info->caching_workers); | |
2233 | ret |= btrfs_start_workers(&fs_info->readahead_workers); | |
2234 | if (ret) { | |
2235 | ret = -ENOMEM; | |
2236 | goto fail_sb_buffer; | |
2237 | } | |
4543df7e | 2238 | |
4575c9cc | 2239 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
2240 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
2241 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 2242 | |
db94535d CM |
2243 | nodesize = btrfs_super_nodesize(disk_super); |
2244 | leafsize = btrfs_super_leafsize(disk_super); | |
2245 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 2246 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
2247 | tree_root->nodesize = nodesize; |
2248 | tree_root->leafsize = leafsize; | |
2249 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 2250 | tree_root->stripesize = stripesize; |
a061fc8d CM |
2251 | |
2252 | sb->s_blocksize = sectorsize; | |
2253 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 2254 | |
39279cc3 CM |
2255 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
2256 | sizeof(disk_super->magic))) { | |
d397712b | 2257 | printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); |
39279cc3 CM |
2258 | goto fail_sb_buffer; |
2259 | } | |
19c00ddc | 2260 | |
941b2ddf KM |
2261 | if (sectorsize < PAGE_SIZE) { |
2262 | printk(KERN_WARNING "btrfs: Incompatible sector size " | |
2263 | "found on %s\n", sb->s_id); | |
2264 | goto fail_sb_buffer; | |
2265 | } | |
2266 | ||
925baedd | 2267 | mutex_lock(&fs_info->chunk_mutex); |
e4404d6e | 2268 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 2269 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 2270 | if (ret) { |
d397712b CM |
2271 | printk(KERN_WARNING "btrfs: failed to read the system " |
2272 | "array on %s\n", sb->s_id); | |
5d4f98a2 | 2273 | goto fail_sb_buffer; |
84eed90f | 2274 | } |
0b86a832 CM |
2275 | |
2276 | blocksize = btrfs_level_size(tree_root, | |
2277 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 2278 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
2279 | |
2280 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2281 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
2282 | ||
2283 | chunk_root->node = read_tree_block(chunk_root, | |
2284 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 2285 | blocksize, generation); |
0b86a832 | 2286 | BUG_ON(!chunk_root->node); |
83121942 DW |
2287 | if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { |
2288 | printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n", | |
2289 | sb->s_id); | |
af31f5e5 | 2290 | goto fail_tree_roots; |
83121942 | 2291 | } |
5d4f98a2 YZ |
2292 | btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); |
2293 | chunk_root->commit_root = btrfs_root_node(chunk_root); | |
0b86a832 | 2294 | |
e17cade2 | 2295 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
d397712b CM |
2296 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), |
2297 | BTRFS_UUID_SIZE); | |
e17cade2 | 2298 | |
0b86a832 | 2299 | ret = btrfs_read_chunk_tree(chunk_root); |
2b82032c | 2300 | if (ret) { |
d397712b CM |
2301 | printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", |
2302 | sb->s_id); | |
af31f5e5 | 2303 | goto fail_tree_roots; |
2b82032c | 2304 | } |
0b86a832 | 2305 | |
dfe25020 CM |
2306 | btrfs_close_extra_devices(fs_devices); |
2307 | ||
af31f5e5 | 2308 | retry_root_backup: |
db94535d CM |
2309 | blocksize = btrfs_level_size(tree_root, |
2310 | btrfs_super_root_level(disk_super)); | |
84234f3a | 2311 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 2312 | |
e20d96d6 | 2313 | tree_root->node = read_tree_block(tree_root, |
db94535d | 2314 | btrfs_super_root(disk_super), |
84234f3a | 2315 | blocksize, generation); |
af31f5e5 CM |
2316 | if (!tree_root->node || |
2317 | !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { | |
83121942 DW |
2318 | printk(KERN_WARNING "btrfs: failed to read tree root on %s\n", |
2319 | sb->s_id); | |
af31f5e5 CM |
2320 | |
2321 | goto recovery_tree_root; | |
83121942 | 2322 | } |
af31f5e5 | 2323 | |
5d4f98a2 YZ |
2324 | btrfs_set_root_node(&tree_root->root_item, tree_root->node); |
2325 | tree_root->commit_root = btrfs_root_node(tree_root); | |
db94535d CM |
2326 | |
2327 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 2328 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 2329 | if (ret) |
af31f5e5 | 2330 | goto recovery_tree_root; |
0b86a832 CM |
2331 | extent_root->track_dirty = 1; |
2332 | ||
2333 | ret = find_and_setup_root(tree_root, fs_info, | |
2334 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
0b86a832 | 2335 | if (ret) |
af31f5e5 | 2336 | goto recovery_tree_root; |
5d4f98a2 | 2337 | dev_root->track_dirty = 1; |
3768f368 | 2338 | |
d20f7043 CM |
2339 | ret = find_and_setup_root(tree_root, fs_info, |
2340 | BTRFS_CSUM_TREE_OBJECTID, csum_root); | |
2341 | if (ret) | |
af31f5e5 | 2342 | goto recovery_tree_root; |
d20f7043 CM |
2343 | |
2344 | csum_root->track_dirty = 1; | |
2345 | ||
8929ecfa YZ |
2346 | fs_info->generation = generation; |
2347 | fs_info->last_trans_committed = generation; | |
8929ecfa | 2348 | |
c59021f8 | 2349 | ret = btrfs_init_space_info(fs_info); |
2350 | if (ret) { | |
2351 | printk(KERN_ERR "Failed to initial space info: %d\n", ret); | |
2352 | goto fail_block_groups; | |
2353 | } | |
2354 | ||
1b1d1f66 JB |
2355 | ret = btrfs_read_block_groups(extent_root); |
2356 | if (ret) { | |
2357 | printk(KERN_ERR "Failed to read block groups: %d\n", ret); | |
2358 | goto fail_block_groups; | |
2359 | } | |
9078a3e1 | 2360 | |
a74a4b97 CM |
2361 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
2362 | "btrfs-cleaner"); | |
57506d50 | 2363 | if (IS_ERR(fs_info->cleaner_kthread)) |
1b1d1f66 | 2364 | goto fail_block_groups; |
a74a4b97 CM |
2365 | |
2366 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
2367 | tree_root, | |
2368 | "btrfs-transaction"); | |
57506d50 | 2369 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 2370 | goto fail_cleaner; |
a74a4b97 | 2371 | |
c289811c CM |
2372 | if (!btrfs_test_opt(tree_root, SSD) && |
2373 | !btrfs_test_opt(tree_root, NOSSD) && | |
2374 | !fs_info->fs_devices->rotating) { | |
2375 | printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " | |
2376 | "mode\n"); | |
2377 | btrfs_set_opt(fs_info->mount_opt, SSD); | |
2378 | } | |
2379 | ||
21adbd5c SB |
2380 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
2381 | if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { | |
2382 | ret = btrfsic_mount(tree_root, fs_devices, | |
2383 | btrfs_test_opt(tree_root, | |
2384 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? | |
2385 | 1 : 0, | |
2386 | fs_info->check_integrity_print_mask); | |
2387 | if (ret) | |
2388 | printk(KERN_WARNING "btrfs: failed to initialize" | |
2389 | " integrity check module %s\n", sb->s_id); | |
2390 | } | |
2391 | #endif | |
2392 | ||
acce952b | 2393 | /* do not make disk changes in broken FS */ |
2394 | if (btrfs_super_log_root(disk_super) != 0 && | |
2395 | !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { | |
e02119d5 CM |
2396 | u64 bytenr = btrfs_super_log_root(disk_super); |
2397 | ||
7c2ca468 | 2398 | if (fs_devices->rw_devices == 0) { |
d397712b CM |
2399 | printk(KERN_WARNING "Btrfs log replay required " |
2400 | "on RO media\n"); | |
7c2ca468 CM |
2401 | err = -EIO; |
2402 | goto fail_trans_kthread; | |
2403 | } | |
e02119d5 CM |
2404 | blocksize = |
2405 | btrfs_level_size(tree_root, | |
2406 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 2407 | |
676e4c86 DC |
2408 | log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); |
2409 | if (!log_tree_root) { | |
2410 | err = -ENOMEM; | |
2411 | goto fail_trans_kthread; | |
2412 | } | |
e02119d5 CM |
2413 | |
2414 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
2415 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
2416 | ||
2417 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
2418 | blocksize, |
2419 | generation + 1); | |
e02119d5 CM |
2420 | ret = btrfs_recover_log_trees(log_tree_root); |
2421 | BUG_ON(ret); | |
e556ce2c YZ |
2422 | |
2423 | if (sb->s_flags & MS_RDONLY) { | |
2424 | ret = btrfs_commit_super(tree_root); | |
2425 | BUG_ON(ret); | |
2426 | } | |
e02119d5 | 2427 | } |
1a40e23b | 2428 | |
76dda93c YZ |
2429 | ret = btrfs_find_orphan_roots(tree_root); |
2430 | BUG_ON(ret); | |
2431 | ||
7c2ca468 | 2432 | if (!(sb->s_flags & MS_RDONLY)) { |
d68fc57b YZ |
2433 | ret = btrfs_cleanup_fs_roots(fs_info); |
2434 | BUG_ON(ret); | |
2435 | ||
5d4f98a2 | 2436 | ret = btrfs_recover_relocation(tree_root); |
d7ce5843 MX |
2437 | if (ret < 0) { |
2438 | printk(KERN_WARNING | |
2439 | "btrfs: failed to recover relocation\n"); | |
2440 | err = -EINVAL; | |
2441 | goto fail_trans_kthread; | |
2442 | } | |
7c2ca468 | 2443 | } |
1a40e23b | 2444 | |
3de4586c CM |
2445 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
2446 | location.type = BTRFS_ROOT_ITEM_KEY; | |
2447 | location.offset = (u64)-1; | |
2448 | ||
3de4586c CM |
2449 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); |
2450 | if (!fs_info->fs_root) | |
7c2ca468 | 2451 | goto fail_trans_kthread; |
3140c9a3 DC |
2452 | if (IS_ERR(fs_info->fs_root)) { |
2453 | err = PTR_ERR(fs_info->fs_root); | |
2454 | goto fail_trans_kthread; | |
2455 | } | |
c289811c | 2456 | |
e3acc2a6 JB |
2457 | if (!(sb->s_flags & MS_RDONLY)) { |
2458 | down_read(&fs_info->cleanup_work_sem); | |
66b4ffd1 JB |
2459 | err = btrfs_orphan_cleanup(fs_info->fs_root); |
2460 | if (!err) | |
2461 | err = btrfs_orphan_cleanup(fs_info->tree_root); | |
e3acc2a6 | 2462 | up_read(&fs_info->cleanup_work_sem); |
59641015 ID |
2463 | |
2464 | if (!err) | |
2465 | err = btrfs_recover_balance(fs_info->tree_root); | |
2466 | ||
66b4ffd1 JB |
2467 | if (err) { |
2468 | close_ctree(tree_root); | |
2469 | return ERR_PTR(err); | |
2470 | } | |
e3acc2a6 JB |
2471 | } |
2472 | ||
0f7d52f4 | 2473 | return tree_root; |
39279cc3 | 2474 | |
7c2ca468 CM |
2475 | fail_trans_kthread: |
2476 | kthread_stop(fs_info->transaction_kthread); | |
3f157a2f | 2477 | fail_cleaner: |
a74a4b97 | 2478 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
2479 | |
2480 | /* | |
2481 | * make sure we're done with the btree inode before we stop our | |
2482 | * kthreads | |
2483 | */ | |
2484 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
2485 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
2486 | ||
1b1d1f66 JB |
2487 | fail_block_groups: |
2488 | btrfs_free_block_groups(fs_info); | |
af31f5e5 CM |
2489 | |
2490 | fail_tree_roots: | |
2491 | free_root_pointers(fs_info, 1); | |
2492 | ||
39279cc3 | 2493 | fail_sb_buffer: |
61d92c32 | 2494 | btrfs_stop_workers(&fs_info->generic_worker); |
306c8b68 | 2495 | btrfs_stop_workers(&fs_info->readahead_workers); |
247e743c | 2496 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2497 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2498 | btrfs_stop_workers(&fs_info->workers); |
2499 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 2500 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 2501 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 2502 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 2503 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 2504 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 2505 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 2506 | btrfs_stop_workers(&fs_info->caching_workers); |
16cdcec7 | 2507 | fail_alloc: |
4543df7e | 2508 | fail_iput: |
586e46e2 ID |
2509 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
2510 | ||
7c2ca468 | 2511 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
4543df7e | 2512 | iput(fs_info->btree_inode); |
ad081f14 | 2513 | fail_bdi: |
7e662854 | 2514 | bdi_destroy(&fs_info->bdi); |
76dda93c YZ |
2515 | fail_srcu: |
2516 | cleanup_srcu_struct(&fs_info->subvol_srcu); | |
7e662854 | 2517 | fail: |
586e46e2 | 2518 | btrfs_close_devices(fs_info->fs_devices); |
6c41761f | 2519 | free_fs_info(fs_info); |
39279cc3 | 2520 | return ERR_PTR(err); |
af31f5e5 CM |
2521 | |
2522 | recovery_tree_root: | |
af31f5e5 CM |
2523 | if (!btrfs_test_opt(tree_root, RECOVERY)) |
2524 | goto fail_tree_roots; | |
2525 | ||
2526 | free_root_pointers(fs_info, 0); | |
2527 | ||
2528 | /* don't use the log in recovery mode, it won't be valid */ | |
2529 | btrfs_set_super_log_root(disk_super, 0); | |
2530 | ||
2531 | /* we can't trust the free space cache either */ | |
2532 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2533 | ||
2534 | ret = next_root_backup(fs_info, fs_info->super_copy, | |
2535 | &num_backups_tried, &backup_index); | |
2536 | if (ret == -1) | |
2537 | goto fail_block_groups; | |
2538 | goto retry_root_backup; | |
eb60ceac CM |
2539 | } |
2540 | ||
f2984462 CM |
2541 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
2542 | { | |
2543 | char b[BDEVNAME_SIZE]; | |
2544 | ||
2545 | if (uptodate) { | |
2546 | set_buffer_uptodate(bh); | |
2547 | } else { | |
7a36ddec | 2548 | printk_ratelimited(KERN_WARNING "lost page write due to " |
f2984462 CM |
2549 | "I/O error on %s\n", |
2550 | bdevname(bh->b_bdev, b)); | |
1259ab75 CM |
2551 | /* note, we dont' set_buffer_write_io_error because we have |
2552 | * our own ways of dealing with the IO errors | |
2553 | */ | |
f2984462 CM |
2554 | clear_buffer_uptodate(bh); |
2555 | } | |
2556 | unlock_buffer(bh); | |
2557 | put_bh(bh); | |
2558 | } | |
2559 | ||
a512bbf8 YZ |
2560 | struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) |
2561 | { | |
2562 | struct buffer_head *bh; | |
2563 | struct buffer_head *latest = NULL; | |
2564 | struct btrfs_super_block *super; | |
2565 | int i; | |
2566 | u64 transid = 0; | |
2567 | u64 bytenr; | |
2568 | ||
2569 | /* we would like to check all the supers, but that would make | |
2570 | * a btrfs mount succeed after a mkfs from a different FS. | |
2571 | * So, we need to add a special mount option to scan for | |
2572 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
2573 | */ | |
2574 | for (i = 0; i < 1; i++) { | |
2575 | bytenr = btrfs_sb_offset(i); | |
2576 | if (bytenr + 4096 >= i_size_read(bdev->bd_inode)) | |
2577 | break; | |
2578 | bh = __bread(bdev, bytenr / 4096, 4096); | |
2579 | if (!bh) | |
2580 | continue; | |
2581 | ||
2582 | super = (struct btrfs_super_block *)bh->b_data; | |
2583 | if (btrfs_super_bytenr(super) != bytenr || | |
2584 | strncmp((char *)(&super->magic), BTRFS_MAGIC, | |
2585 | sizeof(super->magic))) { | |
2586 | brelse(bh); | |
2587 | continue; | |
2588 | } | |
2589 | ||
2590 | if (!latest || btrfs_super_generation(super) > transid) { | |
2591 | brelse(latest); | |
2592 | latest = bh; | |
2593 | transid = btrfs_super_generation(super); | |
2594 | } else { | |
2595 | brelse(bh); | |
2596 | } | |
2597 | } | |
2598 | return latest; | |
2599 | } | |
2600 | ||
4eedeb75 HH |
2601 | /* |
2602 | * this should be called twice, once with wait == 0 and | |
2603 | * once with wait == 1. When wait == 0 is done, all the buffer heads | |
2604 | * we write are pinned. | |
2605 | * | |
2606 | * They are released when wait == 1 is done. | |
2607 | * max_mirrors must be the same for both runs, and it indicates how | |
2608 | * many supers on this one device should be written. | |
2609 | * | |
2610 | * max_mirrors == 0 means to write them all. | |
2611 | */ | |
a512bbf8 YZ |
2612 | static int write_dev_supers(struct btrfs_device *device, |
2613 | struct btrfs_super_block *sb, | |
2614 | int do_barriers, int wait, int max_mirrors) | |
2615 | { | |
2616 | struct buffer_head *bh; | |
2617 | int i; | |
2618 | int ret; | |
2619 | int errors = 0; | |
2620 | u32 crc; | |
2621 | u64 bytenr; | |
a512bbf8 YZ |
2622 | |
2623 | if (max_mirrors == 0) | |
2624 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
2625 | ||
a512bbf8 YZ |
2626 | for (i = 0; i < max_mirrors; i++) { |
2627 | bytenr = btrfs_sb_offset(i); | |
2628 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) | |
2629 | break; | |
2630 | ||
2631 | if (wait) { | |
2632 | bh = __find_get_block(device->bdev, bytenr / 4096, | |
2633 | BTRFS_SUPER_INFO_SIZE); | |
2634 | BUG_ON(!bh); | |
a512bbf8 | 2635 | wait_on_buffer(bh); |
4eedeb75 HH |
2636 | if (!buffer_uptodate(bh)) |
2637 | errors++; | |
2638 | ||
2639 | /* drop our reference */ | |
2640 | brelse(bh); | |
2641 | ||
2642 | /* drop the reference from the wait == 0 run */ | |
2643 | brelse(bh); | |
2644 | continue; | |
a512bbf8 YZ |
2645 | } else { |
2646 | btrfs_set_super_bytenr(sb, bytenr); | |
2647 | ||
2648 | crc = ~(u32)0; | |
2649 | crc = btrfs_csum_data(NULL, (char *)sb + | |
2650 | BTRFS_CSUM_SIZE, crc, | |
2651 | BTRFS_SUPER_INFO_SIZE - | |
2652 | BTRFS_CSUM_SIZE); | |
2653 | btrfs_csum_final(crc, sb->csum); | |
2654 | ||
4eedeb75 HH |
2655 | /* |
2656 | * one reference for us, and we leave it for the | |
2657 | * caller | |
2658 | */ | |
a512bbf8 YZ |
2659 | bh = __getblk(device->bdev, bytenr / 4096, |
2660 | BTRFS_SUPER_INFO_SIZE); | |
2661 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); | |
2662 | ||
4eedeb75 | 2663 | /* one reference for submit_bh */ |
a512bbf8 | 2664 | get_bh(bh); |
4eedeb75 HH |
2665 | |
2666 | set_buffer_uptodate(bh); | |
a512bbf8 YZ |
2667 | lock_buffer(bh); |
2668 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
2669 | } | |
2670 | ||
387125fc CM |
2671 | /* |
2672 | * we fua the first super. The others we allow | |
2673 | * to go down lazy. | |
2674 | */ | |
21adbd5c | 2675 | ret = btrfsic_submit_bh(WRITE_FUA, bh); |
4eedeb75 | 2676 | if (ret) |
a512bbf8 | 2677 | errors++; |
a512bbf8 YZ |
2678 | } |
2679 | return errors < i ? 0 : -1; | |
2680 | } | |
2681 | ||
387125fc CM |
2682 | /* |
2683 | * endio for the write_dev_flush, this will wake anyone waiting | |
2684 | * for the barrier when it is done | |
2685 | */ | |
2686 | static void btrfs_end_empty_barrier(struct bio *bio, int err) | |
2687 | { | |
2688 | if (err) { | |
2689 | if (err == -EOPNOTSUPP) | |
2690 | set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); | |
2691 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
2692 | } | |
2693 | if (bio->bi_private) | |
2694 | complete(bio->bi_private); | |
2695 | bio_put(bio); | |
2696 | } | |
2697 | ||
2698 | /* | |
2699 | * trigger flushes for one the devices. If you pass wait == 0, the flushes are | |
2700 | * sent down. With wait == 1, it waits for the previous flush. | |
2701 | * | |
2702 | * any device where the flush fails with eopnotsupp are flagged as not-barrier | |
2703 | * capable | |
2704 | */ | |
2705 | static int write_dev_flush(struct btrfs_device *device, int wait) | |
2706 | { | |
2707 | struct bio *bio; | |
2708 | int ret = 0; | |
2709 | ||
2710 | if (device->nobarriers) | |
2711 | return 0; | |
2712 | ||
2713 | if (wait) { | |
2714 | bio = device->flush_bio; | |
2715 | if (!bio) | |
2716 | return 0; | |
2717 | ||
2718 | wait_for_completion(&device->flush_wait); | |
2719 | ||
2720 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) { | |
2721 | printk("btrfs: disabling barriers on dev %s\n", | |
2722 | device->name); | |
2723 | device->nobarriers = 1; | |
2724 | } | |
2725 | if (!bio_flagged(bio, BIO_UPTODATE)) { | |
2726 | ret = -EIO; | |
2727 | } | |
2728 | ||
2729 | /* drop the reference from the wait == 0 run */ | |
2730 | bio_put(bio); | |
2731 | device->flush_bio = NULL; | |
2732 | ||
2733 | return ret; | |
2734 | } | |
2735 | ||
2736 | /* | |
2737 | * one reference for us, and we leave it for the | |
2738 | * caller | |
2739 | */ | |
2740 | device->flush_bio = NULL;; | |
2741 | bio = bio_alloc(GFP_NOFS, 0); | |
2742 | if (!bio) | |
2743 | return -ENOMEM; | |
2744 | ||
2745 | bio->bi_end_io = btrfs_end_empty_barrier; | |
2746 | bio->bi_bdev = device->bdev; | |
2747 | init_completion(&device->flush_wait); | |
2748 | bio->bi_private = &device->flush_wait; | |
2749 | device->flush_bio = bio; | |
2750 | ||
2751 | bio_get(bio); | |
21adbd5c | 2752 | btrfsic_submit_bio(WRITE_FLUSH, bio); |
387125fc CM |
2753 | |
2754 | return 0; | |
2755 | } | |
2756 | ||
2757 | /* | |
2758 | * send an empty flush down to each device in parallel, | |
2759 | * then wait for them | |
2760 | */ | |
2761 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
2762 | { | |
2763 | struct list_head *head; | |
2764 | struct btrfs_device *dev; | |
2765 | int errors = 0; | |
2766 | int ret; | |
2767 | ||
2768 | /* send down all the barriers */ | |
2769 | head = &info->fs_devices->devices; | |
2770 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2771 | if (!dev->bdev) { | |
2772 | errors++; | |
2773 | continue; | |
2774 | } | |
2775 | if (!dev->in_fs_metadata || !dev->writeable) | |
2776 | continue; | |
2777 | ||
2778 | ret = write_dev_flush(dev, 0); | |
2779 | if (ret) | |
2780 | errors++; | |
2781 | } | |
2782 | ||
2783 | /* wait for all the barriers */ | |
2784 | list_for_each_entry_rcu(dev, head, dev_list) { | |
2785 | if (!dev->bdev) { | |
2786 | errors++; | |
2787 | continue; | |
2788 | } | |
2789 | if (!dev->in_fs_metadata || !dev->writeable) | |
2790 | continue; | |
2791 | ||
2792 | ret = write_dev_flush(dev, 1); | |
2793 | if (ret) | |
2794 | errors++; | |
2795 | } | |
2796 | if (errors) | |
2797 | return -EIO; | |
2798 | return 0; | |
2799 | } | |
2800 | ||
a512bbf8 | 2801 | int write_all_supers(struct btrfs_root *root, int max_mirrors) |
f2984462 | 2802 | { |
e5e9a520 | 2803 | struct list_head *head; |
f2984462 | 2804 | struct btrfs_device *dev; |
a061fc8d | 2805 | struct btrfs_super_block *sb; |
f2984462 | 2806 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
2807 | int ret; |
2808 | int do_barriers; | |
a236aed1 CM |
2809 | int max_errors; |
2810 | int total_errors = 0; | |
a061fc8d | 2811 | u64 flags; |
f2984462 | 2812 | |
6c41761f | 2813 | max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; |
f2984462 | 2814 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
af31f5e5 | 2815 | backup_super_roots(root->fs_info); |
f2984462 | 2816 | |
6c41761f | 2817 | sb = root->fs_info->super_for_commit; |
a061fc8d | 2818 | dev_item = &sb->dev_item; |
e5e9a520 | 2819 | |
174ba509 | 2820 | mutex_lock(&root->fs_info->fs_devices->device_list_mutex); |
e5e9a520 | 2821 | head = &root->fs_info->fs_devices->devices; |
387125fc CM |
2822 | |
2823 | if (do_barriers) | |
2824 | barrier_all_devices(root->fs_info); | |
2825 | ||
1f78160c | 2826 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2827 | if (!dev->bdev) { |
2828 | total_errors++; | |
2829 | continue; | |
2830 | } | |
2b82032c | 2831 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2832 | continue; |
2833 | ||
2b82032c | 2834 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
2835 | btrfs_set_stack_device_type(dev_item, dev->type); |
2836 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
2837 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
2838 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
2839 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
2840 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
2841 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
2842 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2843 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a512bbf8 | 2844 | |
a061fc8d CM |
2845 | flags = btrfs_super_flags(sb); |
2846 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
2847 | ||
a512bbf8 | 2848 | ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); |
a236aed1 CM |
2849 | if (ret) |
2850 | total_errors++; | |
f2984462 | 2851 | } |
a236aed1 | 2852 | if (total_errors > max_errors) { |
d397712b CM |
2853 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2854 | total_errors); | |
a236aed1 CM |
2855 | BUG(); |
2856 | } | |
f2984462 | 2857 | |
a512bbf8 | 2858 | total_errors = 0; |
1f78160c | 2859 | list_for_each_entry_rcu(dev, head, dev_list) { |
dfe25020 CM |
2860 | if (!dev->bdev) |
2861 | continue; | |
2b82032c | 2862 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
2863 | continue; |
2864 | ||
a512bbf8 YZ |
2865 | ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); |
2866 | if (ret) | |
2867 | total_errors++; | |
f2984462 | 2868 | } |
174ba509 | 2869 | mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); |
a236aed1 | 2870 | if (total_errors > max_errors) { |
d397712b CM |
2871 | printk(KERN_ERR "btrfs: %d errors while writing supers\n", |
2872 | total_errors); | |
a236aed1 CM |
2873 | BUG(); |
2874 | } | |
f2984462 CM |
2875 | return 0; |
2876 | } | |
2877 | ||
a512bbf8 YZ |
2878 | int write_ctree_super(struct btrfs_trans_handle *trans, |
2879 | struct btrfs_root *root, int max_mirrors) | |
eb60ceac | 2880 | { |
e66f709b | 2881 | int ret; |
5f39d397 | 2882 | |
a512bbf8 | 2883 | ret = write_all_supers(root, max_mirrors); |
5f39d397 | 2884 | return ret; |
cfaa7295 CM |
2885 | } |
2886 | ||
5eda7b5e | 2887 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f | 2888 | { |
4df27c4d | 2889 | spin_lock(&fs_info->fs_roots_radix_lock); |
2619ba1f CM |
2890 | radix_tree_delete(&fs_info->fs_roots_radix, |
2891 | (unsigned long)root->root_key.objectid); | |
4df27c4d | 2892 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c YZ |
2893 | |
2894 | if (btrfs_root_refs(&root->root_item) == 0) | |
2895 | synchronize_srcu(&fs_info->subvol_srcu); | |
2896 | ||
581bb050 LZ |
2897 | __btrfs_remove_free_space_cache(root->free_ino_pinned); |
2898 | __btrfs_remove_free_space_cache(root->free_ino_ctl); | |
4df27c4d YZ |
2899 | free_fs_root(root); |
2900 | return 0; | |
2901 | } | |
2902 | ||
2903 | static void free_fs_root(struct btrfs_root *root) | |
2904 | { | |
82d5902d | 2905 | iput(root->cache_inode); |
4df27c4d | 2906 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); |
0ee5dc67 AV |
2907 | if (root->anon_dev) |
2908 | free_anon_bdev(root->anon_dev); | |
4df27c4d YZ |
2909 | free_extent_buffer(root->node); |
2910 | free_extent_buffer(root->commit_root); | |
581bb050 LZ |
2911 | kfree(root->free_ino_ctl); |
2912 | kfree(root->free_ino_pinned); | |
d397712b | 2913 | kfree(root->name); |
2619ba1f | 2914 | kfree(root); |
2619ba1f CM |
2915 | } |
2916 | ||
35b7e476 | 2917 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
2918 | { |
2919 | int ret; | |
2920 | struct btrfs_root *gang[8]; | |
2921 | int i; | |
2922 | ||
76dda93c YZ |
2923 | while (!list_empty(&fs_info->dead_roots)) { |
2924 | gang[0] = list_entry(fs_info->dead_roots.next, | |
2925 | struct btrfs_root, root_list); | |
2926 | list_del(&gang[0]->root_list); | |
2927 | ||
2928 | if (gang[0]->in_radix) { | |
2929 | btrfs_free_fs_root(fs_info, gang[0]); | |
2930 | } else { | |
2931 | free_extent_buffer(gang[0]->node); | |
2932 | free_extent_buffer(gang[0]->commit_root); | |
2933 | kfree(gang[0]); | |
2934 | } | |
2935 | } | |
2936 | ||
d397712b | 2937 | while (1) { |
0f7d52f4 CM |
2938 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, |
2939 | (void **)gang, 0, | |
2940 | ARRAY_SIZE(gang)); | |
2941 | if (!ret) | |
2942 | break; | |
2619ba1f | 2943 | for (i = 0; i < ret; i++) |
5eda7b5e | 2944 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
2945 | } |
2946 | return 0; | |
2947 | } | |
b4100d64 | 2948 | |
c146afad | 2949 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 2950 | { |
c146afad YZ |
2951 | u64 root_objectid = 0; |
2952 | struct btrfs_root *gang[8]; | |
2953 | int i; | |
3768f368 | 2954 | int ret; |
e089f05c | 2955 | |
c146afad YZ |
2956 | while (1) { |
2957 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2958 | (void **)gang, root_objectid, | |
2959 | ARRAY_SIZE(gang)); | |
2960 | if (!ret) | |
2961 | break; | |
5d4f98a2 YZ |
2962 | |
2963 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
c146afad | 2964 | for (i = 0; i < ret; i++) { |
66b4ffd1 JB |
2965 | int err; |
2966 | ||
c146afad | 2967 | root_objectid = gang[i]->root_key.objectid; |
66b4ffd1 JB |
2968 | err = btrfs_orphan_cleanup(gang[i]); |
2969 | if (err) | |
2970 | return err; | |
c146afad YZ |
2971 | } |
2972 | root_objectid++; | |
2973 | } | |
2974 | return 0; | |
2975 | } | |
a2135011 | 2976 | |
c146afad YZ |
2977 | int btrfs_commit_super(struct btrfs_root *root) |
2978 | { | |
2979 | struct btrfs_trans_handle *trans; | |
2980 | int ret; | |
a74a4b97 | 2981 | |
c146afad | 2982 | mutex_lock(&root->fs_info->cleaner_mutex); |
24bbcf04 | 2983 | btrfs_run_delayed_iputs(root); |
a74a4b97 | 2984 | btrfs_clean_old_snapshots(root); |
c146afad | 2985 | mutex_unlock(&root->fs_info->cleaner_mutex); |
c71bf099 YZ |
2986 | |
2987 | /* wait until ongoing cleanup work done */ | |
2988 | down_write(&root->fs_info->cleanup_work_sem); | |
2989 | up_write(&root->fs_info->cleanup_work_sem); | |
2990 | ||
7a7eaa40 | 2991 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2992 | if (IS_ERR(trans)) |
2993 | return PTR_ERR(trans); | |
54aa1f4d | 2994 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2995 | BUG_ON(ret); |
2996 | /* run commit again to drop the original snapshot */ | |
7a7eaa40 | 2997 | trans = btrfs_join_transaction(root); |
3612b495 TI |
2998 | if (IS_ERR(trans)) |
2999 | return PTR_ERR(trans); | |
79154b1b CM |
3000 | btrfs_commit_transaction(trans, root); |
3001 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 3002 | BUG_ON(ret); |
d6bfde87 | 3003 | |
a512bbf8 | 3004 | ret = write_ctree_super(NULL, root, 0); |
c146afad YZ |
3005 | return ret; |
3006 | } | |
3007 | ||
3008 | int close_ctree(struct btrfs_root *root) | |
3009 | { | |
3010 | struct btrfs_fs_info *fs_info = root->fs_info; | |
3011 | int ret; | |
3012 | ||
3013 | fs_info->closing = 1; | |
3014 | smp_mb(); | |
3015 | ||
837d5b6e ID |
3016 | /* pause restriper - we want to resume on mount */ |
3017 | btrfs_pause_balance(root->fs_info); | |
3018 | ||
a2de733c | 3019 | btrfs_scrub_cancel(root); |
4cb5300b CM |
3020 | |
3021 | /* wait for any defraggers to finish */ | |
3022 | wait_event(fs_info->transaction_wait, | |
3023 | (atomic_read(&fs_info->defrag_running) == 0)); | |
3024 | ||
3025 | /* clear out the rbtree of defraggable inodes */ | |
3026 | btrfs_run_defrag_inodes(root->fs_info); | |
3027 | ||
acce952b | 3028 | /* |
3029 | * Here come 2 situations when btrfs is broken to flip readonly: | |
3030 | * | |
3031 | * 1. when btrfs flips readonly somewhere else before | |
3032 | * btrfs_commit_super, sb->s_flags has MS_RDONLY flag, | |
3033 | * and btrfs will skip to write sb directly to keep | |
3034 | * ERROR state on disk. | |
3035 | * | |
3036 | * 2. when btrfs flips readonly just in btrfs_commit_super, | |
ae0e47f0 | 3037 | * and in such case, btrfs cannot write sb via btrfs_commit_super, |
acce952b | 3038 | * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag, |
3039 | * btrfs will cleanup all FS resources first and write sb then. | |
3040 | */ | |
c146afad | 3041 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
acce952b | 3042 | ret = btrfs_commit_super(root); |
3043 | if (ret) | |
3044 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
3045 | } | |
3046 | ||
3047 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
3048 | ret = btrfs_error_commit_super(root); | |
d397712b CM |
3049 | if (ret) |
3050 | printk(KERN_ERR "btrfs: commit super ret %d\n", ret); | |
c146afad | 3051 | } |
0f7d52f4 | 3052 | |
300e4f8a JB |
3053 | btrfs_put_block_group_cache(fs_info); |
3054 | ||
8929ecfa YZ |
3055 | kthread_stop(root->fs_info->transaction_kthread); |
3056 | kthread_stop(root->fs_info->cleaner_kthread); | |
3057 | ||
f25784b3 YZ |
3058 | fs_info->closing = 2; |
3059 | smp_mb(); | |
3060 | ||
b0c68f8b | 3061 | if (fs_info->delalloc_bytes) { |
d397712b | 3062 | printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", |
21380931 | 3063 | (unsigned long long)fs_info->delalloc_bytes); |
b0c68f8b | 3064 | } |
31153d81 | 3065 | if (fs_info->total_ref_cache_size) { |
d397712b CM |
3066 | printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", |
3067 | (unsigned long long)fs_info->total_ref_cache_size); | |
31153d81 | 3068 | } |
bcc63abb | 3069 | |
5d4f98a2 YZ |
3070 | free_extent_buffer(fs_info->extent_root->node); |
3071 | free_extent_buffer(fs_info->extent_root->commit_root); | |
3072 | free_extent_buffer(fs_info->tree_root->node); | |
3073 | free_extent_buffer(fs_info->tree_root->commit_root); | |
3074 | free_extent_buffer(root->fs_info->chunk_root->node); | |
3075 | free_extent_buffer(root->fs_info->chunk_root->commit_root); | |
3076 | free_extent_buffer(root->fs_info->dev_root->node); | |
3077 | free_extent_buffer(root->fs_info->dev_root->commit_root); | |
3078 | free_extent_buffer(root->fs_info->csum_root->node); | |
3079 | free_extent_buffer(root->fs_info->csum_root->commit_root); | |
d20f7043 | 3080 | |
9078a3e1 | 3081 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 3082 | |
c146afad | 3083 | del_fs_roots(fs_info); |
d10c5f31 | 3084 | |
c146afad | 3085 | iput(fs_info->btree_inode); |
9ad6b7bc | 3086 | |
61d92c32 | 3087 | btrfs_stop_workers(&fs_info->generic_worker); |
247e743c | 3088 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 3089 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
3090 | btrfs_stop_workers(&fs_info->workers); |
3091 | btrfs_stop_workers(&fs_info->endio_workers); | |
d20f7043 | 3092 | btrfs_stop_workers(&fs_info->endio_meta_workers); |
cad321ad | 3093 | btrfs_stop_workers(&fs_info->endio_meta_write_workers); |
e6dcd2dc | 3094 | btrfs_stop_workers(&fs_info->endio_write_workers); |
0cb59c99 | 3095 | btrfs_stop_workers(&fs_info->endio_freespace_worker); |
1cc127b5 | 3096 | btrfs_stop_workers(&fs_info->submit_workers); |
16cdcec7 | 3097 | btrfs_stop_workers(&fs_info->delayed_workers); |
bab39bf9 | 3098 | btrfs_stop_workers(&fs_info->caching_workers); |
90519d66 | 3099 | btrfs_stop_workers(&fs_info->readahead_workers); |
d6bfde87 | 3100 | |
21adbd5c SB |
3101 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3102 | if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
3103 | btrfsic_unmount(root, fs_info->fs_devices); | |
3104 | #endif | |
3105 | ||
dfe25020 | 3106 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 3107 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 3108 | |
04160088 | 3109 | bdi_destroy(&fs_info->bdi); |
76dda93c | 3110 | cleanup_srcu_struct(&fs_info->subvol_srcu); |
0b86a832 | 3111 | |
6c41761f | 3112 | free_fs_info(fs_info); |
83a4d548 | 3113 | |
eb60ceac CM |
3114 | return 0; |
3115 | } | |
3116 | ||
1259ab75 | 3117 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 3118 | { |
1259ab75 | 3119 | int ret; |
810191ff | 3120 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 | 3121 | |
2ac55d41 JB |
3122 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf, |
3123 | NULL); | |
1259ab75 CM |
3124 | if (!ret) |
3125 | return ret; | |
3126 | ||
3127 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
3128 | parent_transid); | |
3129 | return !ret; | |
5f39d397 CM |
3130 | } |
3131 | ||
3132 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 3133 | { |
810191ff | 3134 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 3135 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
3136 | buf); |
3137 | } | |
6702ed49 | 3138 | |
5f39d397 CM |
3139 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
3140 | { | |
810191ff | 3141 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
3142 | u64 transid = btrfs_header_generation(buf); |
3143 | struct inode *btree_inode = root->fs_info->btree_inode; | |
b9473439 | 3144 | int was_dirty; |
b4ce94de | 3145 | |
b9447ef8 | 3146 | btrfs_assert_tree_locked(buf); |
ccd467d6 | 3147 | if (transid != root->fs_info->generation) { |
d397712b CM |
3148 | printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " |
3149 | "found %llu running %llu\n", | |
db94535d | 3150 | (unsigned long long)buf->start, |
d397712b CM |
3151 | (unsigned long long)transid, |
3152 | (unsigned long long)root->fs_info->generation); | |
ccd467d6 CM |
3153 | WARN_ON(1); |
3154 | } | |
b9473439 CM |
3155 | was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
3156 | buf); | |
3157 | if (!was_dirty) { | |
3158 | spin_lock(&root->fs_info->delalloc_lock); | |
3159 | root->fs_info->dirty_metadata_bytes += buf->len; | |
3160 | spin_unlock(&root->fs_info->delalloc_lock); | |
3161 | } | |
eb60ceac CM |
3162 | } |
3163 | ||
d3c2fdcf | 3164 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
16cdcec7 MX |
3165 | { |
3166 | /* | |
3167 | * looks as though older kernels can get into trouble with | |
3168 | * this code, they end up stuck in balance_dirty_pages forever | |
3169 | */ | |
3170 | u64 num_dirty; | |
3171 | unsigned long thresh = 32 * 1024 * 1024; | |
3172 | ||
3173 | if (current->flags & PF_MEMALLOC) | |
3174 | return; | |
3175 | ||
3176 | btrfs_balance_delayed_items(root); | |
3177 | ||
3178 | num_dirty = root->fs_info->dirty_metadata_bytes; | |
3179 | ||
3180 | if (num_dirty > thresh) { | |
3181 | balance_dirty_pages_ratelimited_nr( | |
3182 | root->fs_info->btree_inode->i_mapping, 1); | |
3183 | } | |
3184 | return; | |
3185 | } | |
3186 | ||
3187 | void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) | |
35b7e476 | 3188 | { |
188de649 CM |
3189 | /* |
3190 | * looks as though older kernels can get into trouble with | |
3191 | * this code, they end up stuck in balance_dirty_pages forever | |
3192 | */ | |
d6bfde87 | 3193 | u64 num_dirty; |
771ed689 | 3194 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 | 3195 | |
6933c02e | 3196 | if (current->flags & PF_MEMALLOC) |
d6bfde87 CM |
3197 | return; |
3198 | ||
585ad2c3 CM |
3199 | num_dirty = root->fs_info->dirty_metadata_bytes; |
3200 | ||
d6bfde87 CM |
3201 | if (num_dirty > thresh) { |
3202 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 3203 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 3204 | } |
188de649 | 3205 | return; |
35b7e476 | 3206 | } |
6b80053d | 3207 | |
ca7a79ad | 3208 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 3209 | { |
810191ff | 3210 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 3211 | int ret; |
ca7a79ad | 3212 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
d397712b | 3213 | if (ret == 0) |
b4ce94de | 3214 | set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags); |
ce9adaa5 | 3215 | return ret; |
6b80053d | 3216 | } |
0da5468f | 3217 | |
01d658f2 CM |
3218 | static int btree_lock_page_hook(struct page *page, void *data, |
3219 | void (*flush_fn)(void *)) | |
4bef0848 CM |
3220 | { |
3221 | struct inode *inode = page->mapping->host; | |
b9473439 | 3222 | struct btrfs_root *root = BTRFS_I(inode)->root; |
4bef0848 CM |
3223 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
3224 | struct extent_buffer *eb; | |
3225 | unsigned long len; | |
3226 | u64 bytenr = page_offset(page); | |
3227 | ||
3228 | if (page->private == EXTENT_PAGE_PRIVATE) | |
3229 | goto out; | |
3230 | ||
3231 | len = page->private >> 2; | |
f09d1f60 | 3232 | eb = find_extent_buffer(io_tree, bytenr, len); |
4bef0848 CM |
3233 | if (!eb) |
3234 | goto out; | |
3235 | ||
01d658f2 CM |
3236 | if (!btrfs_try_tree_write_lock(eb)) { |
3237 | flush_fn(data); | |
3238 | btrfs_tree_lock(eb); | |
3239 | } | |
4bef0848 | 3240 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); |
b9473439 CM |
3241 | |
3242 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { | |
3243 | spin_lock(&root->fs_info->delalloc_lock); | |
3244 | if (root->fs_info->dirty_metadata_bytes >= eb->len) | |
3245 | root->fs_info->dirty_metadata_bytes -= eb->len; | |
3246 | else | |
3247 | WARN_ON(1); | |
3248 | spin_unlock(&root->fs_info->delalloc_lock); | |
3249 | } | |
3250 | ||
4bef0848 CM |
3251 | btrfs_tree_unlock(eb); |
3252 | free_extent_buffer(eb); | |
3253 | out: | |
01d658f2 CM |
3254 | if (!trylock_page(page)) { |
3255 | flush_fn(data); | |
3256 | lock_page(page); | |
3257 | } | |
4bef0848 CM |
3258 | return 0; |
3259 | } | |
3260 | ||
acce952b | 3261 | static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info, |
3262 | int read_only) | |
3263 | { | |
3264 | if (read_only) | |
3265 | return; | |
3266 | ||
3267 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) | |
3268 | printk(KERN_WARNING "warning: mount fs with errors, " | |
3269 | "running btrfsck is recommended\n"); | |
3270 | } | |
3271 | ||
3272 | int btrfs_error_commit_super(struct btrfs_root *root) | |
3273 | { | |
3274 | int ret; | |
3275 | ||
3276 | mutex_lock(&root->fs_info->cleaner_mutex); | |
3277 | btrfs_run_delayed_iputs(root); | |
3278 | mutex_unlock(&root->fs_info->cleaner_mutex); | |
3279 | ||
3280 | down_write(&root->fs_info->cleanup_work_sem); | |
3281 | up_write(&root->fs_info->cleanup_work_sem); | |
3282 | ||
3283 | /* cleanup FS via transaction */ | |
3284 | btrfs_cleanup_transaction(root); | |
3285 | ||
3286 | ret = write_ctree_super(NULL, root, 0); | |
3287 | ||
3288 | return ret; | |
3289 | } | |
3290 | ||
3291 | static int btrfs_destroy_ordered_operations(struct btrfs_root *root) | |
3292 | { | |
3293 | struct btrfs_inode *btrfs_inode; | |
3294 | struct list_head splice; | |
3295 | ||
3296 | INIT_LIST_HEAD(&splice); | |
3297 | ||
3298 | mutex_lock(&root->fs_info->ordered_operations_mutex); | |
3299 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3300 | ||
3301 | list_splice_init(&root->fs_info->ordered_operations, &splice); | |
3302 | while (!list_empty(&splice)) { | |
3303 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3304 | ordered_operations); | |
3305 | ||
3306 | list_del_init(&btrfs_inode->ordered_operations); | |
3307 | ||
3308 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3309 | } | |
3310 | ||
3311 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3312 | mutex_unlock(&root->fs_info->ordered_operations_mutex); | |
3313 | ||
3314 | return 0; | |
3315 | } | |
3316 | ||
3317 | static int btrfs_destroy_ordered_extents(struct btrfs_root *root) | |
3318 | { | |
3319 | struct list_head splice; | |
3320 | struct btrfs_ordered_extent *ordered; | |
3321 | struct inode *inode; | |
3322 | ||
3323 | INIT_LIST_HEAD(&splice); | |
3324 | ||
3325 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3326 | ||
3327 | list_splice_init(&root->fs_info->ordered_extents, &splice); | |
3328 | while (!list_empty(&splice)) { | |
3329 | ordered = list_entry(splice.next, struct btrfs_ordered_extent, | |
3330 | root_extent_list); | |
3331 | ||
3332 | list_del_init(&ordered->root_extent_list); | |
3333 | atomic_inc(&ordered->refs); | |
3334 | ||
3335 | /* the inode may be getting freed (in sys_unlink path). */ | |
3336 | inode = igrab(ordered->inode); | |
3337 | ||
3338 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3339 | if (inode) | |
3340 | iput(inode); | |
3341 | ||
3342 | atomic_set(&ordered->refs, 1); | |
3343 | btrfs_put_ordered_extent(ordered); | |
3344 | ||
3345 | spin_lock(&root->fs_info->ordered_extent_lock); | |
3346 | } | |
3347 | ||
3348 | spin_unlock(&root->fs_info->ordered_extent_lock); | |
3349 | ||
3350 | return 0; | |
3351 | } | |
3352 | ||
3353 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, | |
3354 | struct btrfs_root *root) | |
3355 | { | |
3356 | struct rb_node *node; | |
3357 | struct btrfs_delayed_ref_root *delayed_refs; | |
3358 | struct btrfs_delayed_ref_node *ref; | |
3359 | int ret = 0; | |
3360 | ||
3361 | delayed_refs = &trans->delayed_refs; | |
3362 | ||
3363 | spin_lock(&delayed_refs->lock); | |
3364 | if (delayed_refs->num_entries == 0) { | |
cfece4db | 3365 | spin_unlock(&delayed_refs->lock); |
acce952b | 3366 | printk(KERN_INFO "delayed_refs has NO entry\n"); |
3367 | return ret; | |
3368 | } | |
3369 | ||
3370 | node = rb_first(&delayed_refs->root); | |
3371 | while (node) { | |
3372 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); | |
3373 | node = rb_next(node); | |
3374 | ||
3375 | ref->in_tree = 0; | |
3376 | rb_erase(&ref->rb_node, &delayed_refs->root); | |
3377 | delayed_refs->num_entries--; | |
3378 | ||
3379 | atomic_set(&ref->refs, 1); | |
3380 | if (btrfs_delayed_ref_is_head(ref)) { | |
3381 | struct btrfs_delayed_ref_head *head; | |
3382 | ||
3383 | head = btrfs_delayed_node_to_head(ref); | |
3384 | mutex_lock(&head->mutex); | |
3385 | kfree(head->extent_op); | |
3386 | delayed_refs->num_heads--; | |
3387 | if (list_empty(&head->cluster)) | |
3388 | delayed_refs->num_heads_ready--; | |
3389 | list_del_init(&head->cluster); | |
3390 | mutex_unlock(&head->mutex); | |
3391 | } | |
3392 | ||
3393 | spin_unlock(&delayed_refs->lock); | |
3394 | btrfs_put_delayed_ref(ref); | |
3395 | ||
3396 | cond_resched(); | |
3397 | spin_lock(&delayed_refs->lock); | |
3398 | } | |
3399 | ||
3400 | spin_unlock(&delayed_refs->lock); | |
3401 | ||
3402 | return ret; | |
3403 | } | |
3404 | ||
3405 | static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t) | |
3406 | { | |
3407 | struct btrfs_pending_snapshot *snapshot; | |
3408 | struct list_head splice; | |
3409 | ||
3410 | INIT_LIST_HEAD(&splice); | |
3411 | ||
3412 | list_splice_init(&t->pending_snapshots, &splice); | |
3413 | ||
3414 | while (!list_empty(&splice)) { | |
3415 | snapshot = list_entry(splice.next, | |
3416 | struct btrfs_pending_snapshot, | |
3417 | list); | |
3418 | ||
3419 | list_del_init(&snapshot->list); | |
3420 | ||
3421 | kfree(snapshot); | |
3422 | } | |
3423 | ||
3424 | return 0; | |
3425 | } | |
3426 | ||
3427 | static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root) | |
3428 | { | |
3429 | struct btrfs_inode *btrfs_inode; | |
3430 | struct list_head splice; | |
3431 | ||
3432 | INIT_LIST_HEAD(&splice); | |
3433 | ||
acce952b | 3434 | spin_lock(&root->fs_info->delalloc_lock); |
5be76758 | 3435 | list_splice_init(&root->fs_info->delalloc_inodes, &splice); |
acce952b | 3436 | |
3437 | while (!list_empty(&splice)) { | |
3438 | btrfs_inode = list_entry(splice.next, struct btrfs_inode, | |
3439 | delalloc_inodes); | |
3440 | ||
3441 | list_del_init(&btrfs_inode->delalloc_inodes); | |
3442 | ||
3443 | btrfs_invalidate_inodes(btrfs_inode->root); | |
3444 | } | |
3445 | ||
3446 | spin_unlock(&root->fs_info->delalloc_lock); | |
3447 | ||
3448 | return 0; | |
3449 | } | |
3450 | ||
3451 | static int btrfs_destroy_marked_extents(struct btrfs_root *root, | |
3452 | struct extent_io_tree *dirty_pages, | |
3453 | int mark) | |
3454 | { | |
3455 | int ret; | |
3456 | struct page *page; | |
3457 | struct inode *btree_inode = root->fs_info->btree_inode; | |
3458 | struct extent_buffer *eb; | |
3459 | u64 start = 0; | |
3460 | u64 end; | |
3461 | u64 offset; | |
3462 | unsigned long index; | |
3463 | ||
3464 | while (1) { | |
3465 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
3466 | mark); | |
3467 | if (ret) | |
3468 | break; | |
3469 | ||
3470 | clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); | |
3471 | while (start <= end) { | |
3472 | index = start >> PAGE_CACHE_SHIFT; | |
3473 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; | |
3474 | page = find_get_page(btree_inode->i_mapping, index); | |
3475 | if (!page) | |
3476 | continue; | |
3477 | offset = page_offset(page); | |
3478 | ||
3479 | spin_lock(&dirty_pages->buffer_lock); | |
3480 | eb = radix_tree_lookup( | |
3481 | &(&BTRFS_I(page->mapping->host)->io_tree)->buffer, | |
3482 | offset >> PAGE_CACHE_SHIFT); | |
3483 | spin_unlock(&dirty_pages->buffer_lock); | |
3484 | if (eb) { | |
3485 | ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY, | |
3486 | &eb->bflags); | |
3487 | atomic_set(&eb->refs, 1); | |
3488 | } | |
3489 | if (PageWriteback(page)) | |
3490 | end_page_writeback(page); | |
3491 | ||
3492 | lock_page(page); | |
3493 | if (PageDirty(page)) { | |
3494 | clear_page_dirty_for_io(page); | |
3495 | spin_lock_irq(&page->mapping->tree_lock); | |
3496 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3497 | page_index(page), | |
3498 | PAGECACHE_TAG_DIRTY); | |
3499 | spin_unlock_irq(&page->mapping->tree_lock); | |
3500 | } | |
3501 | ||
3502 | page->mapping->a_ops->invalidatepage(page, 0); | |
3503 | unlock_page(page); | |
3504 | } | |
3505 | } | |
3506 | ||
3507 | return ret; | |
3508 | } | |
3509 | ||
3510 | static int btrfs_destroy_pinned_extent(struct btrfs_root *root, | |
3511 | struct extent_io_tree *pinned_extents) | |
3512 | { | |
3513 | struct extent_io_tree *unpin; | |
3514 | u64 start; | |
3515 | u64 end; | |
3516 | int ret; | |
3517 | ||
3518 | unpin = pinned_extents; | |
3519 | while (1) { | |
3520 | ret = find_first_extent_bit(unpin, 0, &start, &end, | |
3521 | EXTENT_DIRTY); | |
3522 | if (ret) | |
3523 | break; | |
3524 | ||
3525 | /* opt_discard */ | |
5378e607 LD |
3526 | if (btrfs_test_opt(root, DISCARD)) |
3527 | ret = btrfs_error_discard_extent(root, start, | |
3528 | end + 1 - start, | |
3529 | NULL); | |
acce952b | 3530 | |
3531 | clear_extent_dirty(unpin, start, end, GFP_NOFS); | |
3532 | btrfs_error_unpin_extent_range(root, start, end); | |
3533 | cond_resched(); | |
3534 | } | |
3535 | ||
3536 | return 0; | |
3537 | } | |
3538 | ||
3539 | static int btrfs_cleanup_transaction(struct btrfs_root *root) | |
3540 | { | |
3541 | struct btrfs_transaction *t; | |
3542 | LIST_HEAD(list); | |
3543 | ||
3544 | WARN_ON(1); | |
3545 | ||
acce952b | 3546 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
3547 | ||
a4abeea4 | 3548 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3549 | list_splice_init(&root->fs_info->trans_list, &list); |
a4abeea4 JB |
3550 | root->fs_info->trans_no_join = 1; |
3551 | spin_unlock(&root->fs_info->trans_lock); | |
3552 | ||
acce952b | 3553 | while (!list_empty(&list)) { |
3554 | t = list_entry(list.next, struct btrfs_transaction, list); | |
3555 | if (!t) | |
3556 | break; | |
3557 | ||
3558 | btrfs_destroy_ordered_operations(root); | |
3559 | ||
3560 | btrfs_destroy_ordered_extents(root); | |
3561 | ||
3562 | btrfs_destroy_delayed_refs(t, root); | |
3563 | ||
3564 | btrfs_block_rsv_release(root, | |
3565 | &root->fs_info->trans_block_rsv, | |
3566 | t->dirty_pages.dirty_bytes); | |
3567 | ||
3568 | /* FIXME: cleanup wait for commit */ | |
3569 | t->in_commit = 1; | |
3570 | t->blocked = 1; | |
3571 | if (waitqueue_active(&root->fs_info->transaction_blocked_wait)) | |
3572 | wake_up(&root->fs_info->transaction_blocked_wait); | |
3573 | ||
3574 | t->blocked = 0; | |
3575 | if (waitqueue_active(&root->fs_info->transaction_wait)) | |
3576 | wake_up(&root->fs_info->transaction_wait); | |
acce952b | 3577 | |
acce952b | 3578 | t->commit_done = 1; |
3579 | if (waitqueue_active(&t->commit_wait)) | |
3580 | wake_up(&t->commit_wait); | |
acce952b | 3581 | |
3582 | btrfs_destroy_pending_snapshots(t); | |
3583 | ||
3584 | btrfs_destroy_delalloc_inodes(root); | |
3585 | ||
a4abeea4 | 3586 | spin_lock(&root->fs_info->trans_lock); |
acce952b | 3587 | root->fs_info->running_transaction = NULL; |
a4abeea4 | 3588 | spin_unlock(&root->fs_info->trans_lock); |
acce952b | 3589 | |
3590 | btrfs_destroy_marked_extents(root, &t->dirty_pages, | |
3591 | EXTENT_DIRTY); | |
3592 | ||
3593 | btrfs_destroy_pinned_extent(root, | |
3594 | root->fs_info->pinned_extents); | |
3595 | ||
13c5a93e | 3596 | atomic_set(&t->use_count, 0); |
acce952b | 3597 | list_del_init(&t->list); |
3598 | memset(t, 0, sizeof(*t)); | |
3599 | kmem_cache_free(btrfs_transaction_cachep, t); | |
3600 | } | |
3601 | ||
a4abeea4 JB |
3602 | spin_lock(&root->fs_info->trans_lock); |
3603 | root->fs_info->trans_no_join = 0; | |
3604 | spin_unlock(&root->fs_info->trans_lock); | |
acce952b | 3605 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); |
acce952b | 3606 | |
3607 | return 0; | |
3608 | } | |
3609 | ||
d1310b2e | 3610 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 3611 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 3612 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
4bb31e92 | 3613 | .readpage_io_failed_hook = btree_io_failed_hook, |
0b86a832 | 3614 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
3615 | /* note we're sharing with inode.c for the merge bio hook */ |
3616 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 3617 | }; |